{"data_through":"2026-08-17","ranking":{"aggregation":"combined_mean_at_n","compute_normalized":false,"metric":"Clean accuracy over 189 retained questions","ordering":"descending exact rational Clean score","protocol_interpretation":"mean@N is the arithmetic mean across N attempts; it is not best-of-N, pass@N, or majority voting.","tie_policy":"competition exact-rational ties"},"release_id":"gpqa-clean-results-v0.1.0","rows":[{"attempts_per_question":7,"clean":{"denominator":189,"numerator":184,"percent":"97.35%"},"clean_question_n":189,"configuration_date":"2026-04-24","configuration_id":"gpqa_cfg_v1_7934de713f82e39e9592c19190f2c0c4b7a5afc288e8b146a975a0b4243d3076","delta":{"denominator":2079,"numerator":71,"percentage_points":"+3.42 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_d0053cf20fd93eaa5868c7700e33fef962ac314e29c59e42d96b68879014d4d7","evidence_class":"derived_uniform","evidence_variant_id":"gpqa_evd_v1_a94afa1168cdc0773d4dda8f253740979e77cf3cf54f55f8aa8ef4a601d5ad75","model_configuration":"gpt-5.5-pro-pre-release_xhigh","observed_competition_rank":1,"organization":"OpenAI","original":{"denominator":33,"numerator":31,"percent":"93.94%"},"original_question_n":198,"protocol":"mean@7","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"derived_uniform_epoch_subset"},{"attempts_per_question":1,"clean":{"denominator":189,"numerator":184,"percent":"97.35%"},"clean_question_n":189,"configuration_date":"2026-08-17","configuration_id":"gpqa_cfg_v1_bfdb9cba805be9f524447f939b0d26c6c2a84b2eb93aa007c9f5c6d024b1fd8a","delta":{"denominator":2079,"numerator":71,"percentage_points":"+3.42 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_223fdbc5d16a5cb48e4f360d845c8ae5407a47374e8f7513e71d9a1ce2ac0836","evidence_class":"project_native","evidence_variant_id":"gpqa_evd_v1_26554676aa4c49359991cc9f9b408328e46c02a13be963472c8e02f194af8d26","model_configuration":"Gemini 3.6 Flash (high)","observed_competition_rank":1,"organization":"Google","original":{"denominator":33,"numerator":31,"percent":"93.94%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"complete_with_transport_recovery"},{"attempts_per_question":8,"clean":{"denominator":36,"numerator":35,"percent":"97.22%"},"clean_question_n":189,"configuration_date":"2026-04-24","configuration_id":"gpqa_cfg_v1_524a33f2416077d9ee7c8a56c4e245d2737cfc16dc1505dcc72a7b1994afc88a","delta":{"denominator":528,"numerator":17,"percentage_points":"+3.22 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_797dcb2ae3c120447661d4ccc2860d516d7f0155413c03816d86c43ba4f17ae0","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_248b338d575cfa178b0122eaa86cad889196454926526426ee594ad104499afd","model_configuration":"gpt-5.5-pre-release_xhigh","observed_competition_rank":3,"organization":"OpenAI","original":{"denominator":1584,"numerator":1489,"percent":"94.00%"},"original_question_n":198,"protocol":"mean@8","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":8,"clean":{"denominator":168,"numerator":163,"percent":"97.02%"},"clean_question_n":189,"configuration_date":"2026-08-16","configuration_id":"gpqa_cfg_v1_c0291f9a5fc3b09556c001225b7f3d6b3b1de8db4895ffa3e136dbc321bb3809","delta":{"denominator":1232,"numerator":31,"percentage_points":"+2.52 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_9a3d010c3218fc3c7862a01c0ba72a02508f075c4f5bf3d85d2f71765f4cdcce","evidence_class":"targeted_recovery","evidence_variant_id":"gpqa_evd_v1_49152866a50817f6fe994c5543da3fb80dd1e8349a147f155d18b9c761ac27e7","model_configuration":"gemini-3.1-pro-preview","observed_competition_rank":4,"organization":"Google DeepMind","original":{"denominator":528,"numerator":499,"percent":"94.51%"},"original_question_n":198,"protocol":"mean@8","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"complete_with_epoch_targeted_transport_recovery"},{"attempts_per_question":1,"clean":{"denominator":63,"numerator":61,"percent":"96.83%"},"clean_question_n":189,"configuration_date":"2026-08-17","configuration_id":"gpqa_cfg_v1_e99e75b593bbdcaf766b1b732e2a1f92ac6b0acbdc8bd59190ce0972000f26e1","delta":{"denominator":1386,"numerator":47,"percentage_points":"+3.39 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_dc5be65670991d694c17df37c817319ec24167d10e9fbd23aa16c4fec6d522d6","evidence_class":"project_native","evidence_variant_id":"gpqa_evd_v1_c05840e26a572eb38e857946454ba564f6a5b66e80cfd79471bbc81e648ae354","model_configuration":"GPT-5.6 Sol (max)","observed_competition_rank":5,"organization":"OpenAI","original":{"denominator":198,"numerator":185,"percent":"93.43%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_all_normal"},{"attempts_per_question":8,"clean":{"denominator":756,"numerator":731,"percent":"96.69%"},"clean_question_n":189,"configuration_date":"2026-08-16","configuration_id":"gpqa_cfg_v1_2f4d76e7d0e2db97f02626aaf55d382c339a5f505c3d382cb8492f9239d5894c","delta":{"denominator":33264,"numerator":1147,"percentage_points":"+3.45 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_b7cc1b2197543abe6f081a69f7b4b32fb658d1845a8a3af69f8d202cfdf1237a","evidence_class":"targeted_recovery","evidence_variant_id":"gpqa_evd_v1_8d236b9fb51cb9b3b3c96cd1cca3a77a846671644b720ba0ea833a30af54db86","model_configuration":"gpt-5.4-2026-03-05_xhigh","observed_competition_rank":6,"organization":"OpenAI","original":{"denominator":1584,"numerator":1477,"percent":"93.24%"},"original_question_n":198,"protocol":"mean@8","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"complete_with_epoch_targeted_transport_recovery"},{"attempts_per_question":1,"clean":{"denominator":27,"numerator":26,"percent":"96.30%"},"clean_question_n":189,"configuration_date":"2026-08-06","configuration_id":"gpqa_cfg_v1_283043bba1e8117ffaa1f9aace7567853f92188831fd41efb8f9da8fb29f2f3f","delta":{"denominator":54,"numerator":1,"percentage_points":"+1.85 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_f340a350d20d1656f653d6ccc29e95473ba7ed10f7be6c073ed7e5162f3b724c","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_98be177c252ba543c42683d01cf8c78b74a9f9260b1fc9f67ef99db0501a0fb0","model_configuration":"gemini-3.1-pro-preview_high","observed_competition_rank":7,"organization":"Google DeepMind","original":{"denominator":18,"numerator":17,"percent":"94.44%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":8,"clean":{"denominator":1512,"numerator":1453,"percent":"96.10%"},"clean_question_n":189,"configuration_date":"2026-05-22","configuration_id":"gpqa_cfg_v1_c61eab9033f8b5854c5649db1e12ecba8df925dfd4b6c0455cbe678320b35d78","delta":{"denominator":4158,"numerator":137,"percentage_points":"+3.29 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_46db62f2e515b497fc8911933a8451454331ed63da7531543e40384debed32d7","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_31e2d1edf2d19e00df42190b15321c5b234078a1744c36f57f56815a58d311b8","model_configuration":"gemini-3.5-flash_high","observed_competition_rank":8,"organization":"Google","original":{"denominator":264,"numerator":245,"percent":"92.80%"},"original_question_n":198,"protocol":"mean@8","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":1,"clean":{"denominator":189,"numerator":181,"percent":"95.77%"},"clean_question_n":189,"configuration_date":"2026-08-17","configuration_id":"gpqa_cfg_v1_02806112b3fe0147c940db998df739ae579b592620fa54621d718d9ceef20010","delta":{"denominator":4158,"numerator":139,"percentage_points":"+3.34 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_cb8f2ea222449667ac21c1d4c90b2900c2ef60423d279cbdf95dad2b69a10da9","evidence_class":"project_native","evidence_variant_id":"gpqa_evd_v1_7e75c86ea6628a8b743dede137652fad934dfc638af5c3b3141848607ff5cb29","model_configuration":"Grok 4.6 (high)","observed_competition_rank":9,"organization":"xAI","original":{"denominator":66,"numerator":61,"percent":"92.42%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_all_normal"},{"attempts_per_question":1,"clean":{"denominator":189,"numerator":181,"percent":"95.77%"},"clean_question_n":189,"configuration_date":"2026-08-17","configuration_id":"gpqa_cfg_v1_86c15018f283c868987b63ef745efbca663fee38b52cc60a94981c8efd381fab","delta":{"denominator":2079,"numerator":59,"percentage_points":"+2.84 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_ea1888a35618a235b72c73014b2b8a255e891f289fe2e9f58127eab6738b782c","evidence_class":"project_native","evidence_variant_id":"gpqa_evd_v1_6bc2ac5f0c651c7dcfe59063ccdfdaa19bc85d3a1340e90e42815b766a7820ed","model_configuration":"DeepSeek V4 Flash (max)","observed_competition_rank":9,"organization":"DeepSeek","original":{"denominator":99,"numerator":92,"percent":"92.93%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_all_normal"},{"attempts_per_question":8,"clean":{"denominator":378,"numerator":361,"percent":"95.50%"},"clean_question_n":189,"configuration_date":"2025-11-19","configuration_id":"gpqa_cfg_v1_75b5c82ad8fa3f7520d6c30258334fd2da432a999b7c23be7a10a04ab6bfe1de","delta":{"denominator":33264,"numerator":961,"percentage_points":"+2.89 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_061c2f393ff622d4e751ae0e08d01b67d4af4b313eff6a92d9781a5c11104f22","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_8400f0f288172f70235d3c08757a5ec0e10f3de0859cdf05109f4832d65145bd","model_configuration":"gemini-3-pro-preview","observed_competition_rank":11,"organization":"Google DeepMind","original":{"denominator":176,"numerator":163,"percent":"92.61%"},"original_question_n":198,"protocol":"mean@8","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":1,"clean":{"denominator":21,"numerator":20,"percent":"95.24%"},"clean_question_n":189,"configuration_date":"2026-08-17","configuration_id":"gpqa_cfg_v1_a30430189e9844037ecd5dd092ea5d76f41c830802306d85f70843d9e087f3cd","delta":{"denominator":462,"numerator":13,"percentage_points":"+2.81 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_53d27296521152fec647a546e2bae59f8d0b62d0db3fc9dfe258ea0fb3a9a762","evidence_class":"project_native","evidence_variant_id":"gpqa_evd_v1_ed610c61828b46c64fc27e91d164993b3f17894dbe48858e76abf3048a53c522","model_configuration":"DeepSeek V4 Pro (max)","observed_competition_rank":12,"organization":"DeepSeek","original":{"denominator":66,"numerator":61,"percent":"92.42%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_all_normal"},{"attempts_per_question":1,"clean":{"denominator":21,"numerator":20,"percent":"95.24%"},"clean_question_n":189,"configuration_date":"2026-08-06","configuration_id":"gpqa_cfg_v1_a4204d1bf0943d754d0da9cb0991f909d6a74aecfad020560a30706bf493ea5d","delta":{"denominator":693,"numerator":16,"percentage_points":"+2.31 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_4a3ce3ae867ebf1571567fbd514c98dabb03030370b51bf2cda912beee66f06d","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_fbdcb2eda572d6c2000e7447375499cee1edb211e4db200e6c985e09f29cab8d","model_configuration":"claude-opus-5","observed_competition_rank":12,"organization":"Anthropic","original":{"denominator":99,"numerator":92,"percent":"92.93%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":8,"clean":{"denominator":504,"numerator":479,"percent":"95.04%"},"clean_question_n":189,"configuration_date":"2025-12-13","configuration_id":"gpqa_cfg_v1_0101b0f415adc0308a5aeb576bd938ac45457f4a2b41b2f0ec765daf4da164d7","delta":{"denominator":1848,"numerator":67,"percentage_points":"+3.63 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_aa060b8f563068bef04d82a383a85c2b89b4a1e4df2ea9536d763f3048f49a7e","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_24dac4fd7128a97a3ed42af765c34b8cfe6575cf7653b97c34a6784dcb7f10e8","model_configuration":"gpt-5.2-2025-12-11_xhigh","observed_competition_rank":14,"organization":"OpenAI","original":{"denominator":198,"numerator":181,"percent":"91.41%"},"original_question_n":198,"protocol":"mean@8","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":1,"clean":{"denominator":189,"numerator":179,"percent":"94.71%"},"clean_question_n":189,"configuration_date":"2026-08-17","configuration_id":"gpqa_cfg_v1_211a09311a1e6c34a2d98ce45ee46c8aab95d0380c75c39c262f0d5dd6392e90","delta":{"denominator":4158,"numerator":95,"percentage_points":"+2.28 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_b9b889eeac2e0aa50f5bbb745478dcc5364150d9463dbd3c0d87da422abca143","evidence_class":"project_native","evidence_variant_id":"gpqa_evd_v1_8c7c047ce4c5007a73283a6af356345a69e2810479df69b799ca9741cf9d627b","model_configuration":"GPT-5.6 Luna (max)","observed_competition_rank":15,"organization":"OpenAI","original":{"denominator":66,"numerator":61,"percent":"92.42%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_all_normal"},{"attempts_per_question":1,"clean":{"denominator":189,"numerator":179,"percent":"94.71%"},"clean_question_n":189,"configuration_date":"2026-08-07","configuration_id":"gpqa_cfg_v1_bc3e18e0e130722c5f1e133ef0ec99c49ebc62c0cf1d9a48f4491777870f83b6","delta":{"denominator":2079,"numerator":79,"percentage_points":"+3.80 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_ad363578f65b2e9b208a3f95847b49869248d41bd9700e746dfd783640e4ac7c","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_202f68076109b3a6baa05665805091c8c81bcdae256914acbf9aab584d424f67","model_configuration":"qwen3.7-max","observed_competition_rank":15,"organization":"Alibaba","original":{"denominator":11,"numerator":10,"percent":"90.91%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":1,"clean":{"denominator":189,"numerator":179,"percent":"94.71%"},"clean_question_n":189,"configuration_date":"2026-08-07","configuration_id":"gpqa_cfg_v1_c1662dd832d707580c960ccdf5da02a7a38f399834c024fd516e6015ff35eaa2","delta":{"denominator":2079,"numerator":58,"percentage_points":"+2.79 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_6dbe971b027e9b7ff9f5a2ac758872199004bc23eba046bb852d8762b12a2fbb","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_b2979d2667e6d1608bb2130cf65b9073f482e281753982f55d3545e6c324a4d2","model_configuration":"kimi-k3_high","observed_competition_rank":15,"organization":"Moonshot","original":{"denominator":99,"numerator":91,"percent":"91.92%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":8,"clean":{"denominator":504,"numerator":473,"percent":"93.85%"},"clean_question_n":189,"configuration_date":"2026-05-05","configuration_id":"gpqa_cfg_v1_74454bd1c2bf40c9cfef8af6222ded9c2c33796a97bb3e76f4227ab403ab23f6","delta":{"denominator":1848,"numerator":59,"percentage_points":"+3.19 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_6c1185f91a27e0422602ee91305797592531d5e0ea144e8c4368cf375c360fdb","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_0de7b89e2b44977dab0e6ffc0914fa006f3ae8b20f8113c181a6fa44f9281409","model_configuration":"gpt-5.5_low","observed_competition_rank":18,"organization":"OpenAI","original":{"denominator":396,"numerator":359,"percent":"90.66%"},"original_question_n":198,"protocol":"mean@8","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":6,"clean":{"denominator":81,"numerator":76,"percent":"93.83%"},"clean_question_n":189,"configuration_date":"2026-05-01","configuration_id":"gpqa_cfg_v1_555ebc09735857529c6e51f2975f27f05313997eae5072969000efc2baf8faa7","delta":{"denominator":891,"numerator":26,"percentage_points":"+2.92 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_8a99bf54406f9c83dff96771b325f108b7d26ac67c34b6fbe29422becea78312","evidence_class":"derived_uniform","evidence_variant_id":"gpqa_evd_v1_ad73cf533c8248b403ff3591bd404a3bc2f9a571e2750b18265e87582ac8cf2a","model_configuration":"kimi-k2.6","observed_competition_rank":19,"organization":"Moonshot","original":{"denominator":11,"numerator":10,"percent":"90.91%"},"original_question_n":198,"protocol":"mean@6","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"derived_uniform_epoch_subset"},{"attempts_per_question":1,"clean":{"denominator":63,"numerator":59,"percent":"93.65%"},"clean_question_n":189,"configuration_date":"2026-08-17","configuration_id":"gpqa_cfg_v1_0acf0885980bdc183084c5deb8f6341eda2074c91344c8d89bebce44801f4394","delta":{"denominator":693,"numerator":19,"percentage_points":"+2.74 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_17a8008accd99ed44e41a2068583eb613fb01e9f0327b7fd4f50db4a94c9ea26","evidence_class":"project_native","evidence_variant_id":"gpqa_evd_v1_0abc2af9bad5ea0ddf358b460cf0e986e41f057ec160ae64237b3bdf3c239496","model_configuration":"GPT-5.6 Terra (max)","observed_competition_rank":20,"organization":"OpenAI","original":{"denominator":11,"numerator":10,"percent":"90.91%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_all_normal"},{"attempts_per_question":1,"clean":{"denominator":63,"numerator":59,"percent":"93.65%"},"clean_question_n":189,"configuration_date":"2026-08-10","configuration_id":"gpqa_cfg_v1_3209319a9e249f3a03cbe28de9fcd90b9c603d24c23e54f4d19b3acf0eaa894d","delta":{"denominator":693,"numerator":19,"percentage_points":"+2.74 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_a130b7979f29bc1b9d879224b42a693a1044834fe43610fa05909a86b0808385","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_e16b10917c76b9a9f0fe61840b40054fd59d27d4a8f1b0ec979013a0a244091d","model_configuration":"MiniMax-M3","observed_competition_rank":20,"organization":"MiniMax","original":{"denominator":11,"numerator":10,"percent":"90.91%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":1,"clean":{"denominator":63,"numerator":59,"percent":"93.65%"},"clean_question_n":189,"configuration_date":"2026-07-15","configuration_id":"gpqa_cfg_v1_439fcb4bf6820e10b1024c2303f377da9ff1a1106a368e9fbe08acd16c8f32f9","delta":{"denominator":693,"numerator":26,"percentage_points":"+3.75 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_8118b09fe4e06e98040342560692004b81a072d57e690453ac4b706d0cb591b3","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_993a52cea93d40f3ea6090f1ccfc20c3fd1b37331197eda696eff23dfbcc76f4","model_configuration":"gpt-5.4-2026-03-05_high","observed_competition_rank":20,"organization":"OpenAI","original":{"denominator":99,"numerator":89,"percent":"89.90%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":1,"clean":{"denominator":63,"numerator":59,"percent":"93.65%"},"clean_question_n":189,"configuration_date":"2026-08-17","configuration_id":"gpqa_cfg_v1_5af397530c003e894fc8b0e8c6dfa9efeb6966457911dae825b08a1198fbb480","delta":{"denominator":154,"numerator":5,"percentage_points":"+3.25 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_de2a82fcc1678230a76f2d08226dab25e981b5e3b13ed5a4d704ea96b0f13c93","evidence_class":"project_native","evidence_variant_id":"gpqa_evd_v1_24901ab213808f65c6265c08123405c0d8fa99e9ee3f0cec372e5734be70cd37","model_configuration":"DeepSeek V4 Flash (high)","observed_competition_rank":20,"organization":"DeepSeek","original":{"denominator":198,"numerator":179,"percent":"90.40%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"complete_with_provider_terminal_incorrect"},{"attempts_per_question":8,"clean":{"denominator":63,"numerator":59,"percent":"93.65%"},"clean_question_n":189,"configuration_date":"2026-04-17","configuration_id":"gpqa_cfg_v1_6fb295e570562ea20a65bcb23229741b5939aa34be0bd746777e7ba4407013f1","delta":{"denominator":2772,"numerator":97,"percentage_points":"+3.50 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_e2b45798d741d675bbda1d8e81f0a3375c43f07e431ec8c8cedede4b33a6b8db","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_5f719566029f9039695afd7fa3022bab71168d50ea7d1f7e7bce08d69e1e2143","model_configuration":"claude-opus-4-7_xhigh","observed_competition_rank":20,"organization":"Anthropic","original":{"denominator":132,"numerator":119,"percent":"90.15%"},"original_question_n":198,"protocol":"mean@8","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":1,"clean":{"denominator":63,"numerator":59,"percent":"93.65%"},"clean_question_n":189,"configuration_date":"2026-08-07","configuration_id":"gpqa_cfg_v1_84d05ad544dcd1793dbdc24f941a24829266c8784cb6aeac8d0b08838b5b63f4","delta":{"denominator":693,"numerator":26,"percentage_points":"+3.75 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_694d453f03f600041e46e00ce0428ffb49531b0c37a69594aaa5288ba588fb6c","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_f0076ff82b55cc6aa5ba6b90b2b2bb73c109e21b23b3b83961abf972c178d0c9","model_configuration":"gpt-5.6-sol_low","observed_competition_rank":20,"organization":"OpenAI","original":{"denominator":99,"numerator":89,"percent":"89.90%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":8,"clean":{"denominator":63,"numerator":59,"percent":"93.65%"},"clean_question_n":189,"configuration_date":"2026-02-06","configuration_id":"gpqa_cfg_v1_d1854f026c5b433ca83c02a7d0582332f2eb4594f2a0f3223a075ac97db14769","delta":{"denominator":5544,"numerator":173,"percentage_points":"+3.12 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_a1d0d683a699d85732a54ec4ce6fe93f60323ce13c14f2558ff6922042da7709","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_38d18632501697af4ada9c62e750c8fd8b8cc0901b5dd7425f68c740790779b1","model_configuration":"claude-opus-4-6_32K","observed_competition_rank":20,"organization":"Anthropic","original":{"denominator":264,"numerator":239,"percent":"90.53%"},"original_question_n":198,"protocol":"mean@8","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":1,"clean":{"denominator":189,"numerator":176,"percent":"93.12%"},"clean_question_n":189,"configuration_date":"2026-08-10","configuration_id":"gpqa_cfg_v1_17d1971555900db9f02b08fe2063bb81c1f406eb3f99ded57d6366a966344171","delta":{"denominator":2079,"numerator":67,"percentage_points":"+3.22 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_eed4e81138caaf681f745732a4045a78eddc7f7c37bc6f04fad9417e2024140e","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_564911d072eaf17aab93846ce51f5c9aa06b468e0e1dc1a8e86bdeb3b5db0502","model_configuration":"glm-5.1","observed_competition_rank":27,"organization":"Z.ai (Zhipu AI)","original":{"denominator":99,"numerator":89,"percent":"89.90%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":1,"clean":{"denominator":189,"numerator":176,"percent":"93.12%"},"clean_question_n":189,"configuration_date":"2026-08-07","configuration_id":"gpqa_cfg_v1_e0339f47a0e11c791da4e131cf06ead62b4fe9b676d100b883e92ae30cec47ad","delta":{"denominator":2079,"numerator":46,"percentage_points":"+2.21 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_e2e8c76ceab18cf5dc14a36af0f055bed47763cb11426d5b0ac65a54a57055fb","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_8607afe2336d204b79bed741f39ffa46a59fc049b7bc0649101afacead6b91e9","model_configuration":"deepseek-v4-pro_high","observed_competition_rank":27,"organization":"DeepSeek","original":{"denominator":11,"numerator":10,"percent":"90.91%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":1,"clean":{"denominator":27,"numerator":25,"percent":"92.59%"},"clean_question_n":189,"configuration_date":"2026-08-06","configuration_id":"gpqa_cfg_v1_319d02deeba5eff1fe5e7b0068499b112d9a54efc9456189fc1777dbacdee7a6","delta":{"denominator":27,"numerator":1,"percentage_points":"+3.70 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_3ff6adeba5ec1314c7481630677123f787237af0ef3d9c0b7172929cc481810e","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_f53d712a85d9dccef1d05e28777cb8ad3de3ff8e247d7446ec4170734ed60ef5","model_configuration":"gemini-3.5-flash_low","observed_competition_rank":29,"organization":"Google","original":{"denominator":9,"numerator":8,"percent":"88.89%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":1,"clean":{"denominator":63,"numerator":58,"percent":"92.06%"},"clean_question_n":189,"configuration_date":"2026-07-15","configuration_id":"gpqa_cfg_v1_1e7553c1137d1a896e5336d0f9e36f600eb3f8c1981c9496e39923a6dceac58a","delta":{"denominator":63,"numerator":2,"percentage_points":"+3.17 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_a969aad50a7ffba1c59209fd68db8a0d8ef59fb0d2a1724ee6e52c351c4d87eb","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_f5d2a74fcae956da52dad90eff78f31de476b4545ec1e0fe6e3bc93288edd280","model_configuration":"gpt-5.4-2026-03-05_medium","observed_competition_rank":30,"organization":"OpenAI","original":{"denominator":9,"numerator":8,"percent":"88.89%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":1,"clean":{"denominator":63,"numerator":58,"percent":"92.06%"},"clean_question_n":189,"configuration_date":"2026-08-06","configuration_id":"gpqa_cfg_v1_c09343b2de7a3b30deceaaba8fcf1b1e2930f59ccdd6ba4efcd34c200b253701","delta":{"denominator":1386,"numerator":37,"percentage_points":"+2.67 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_7c5f5aedd5059e41d6225ac7422c2245780ae7ff7ff62dc0027bbae0edb76c98","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_43fd53c4f6f37ac29010b1ec216659d5796e7d85c5a296a0cfde43c8f9e6932b","model_configuration":"gemini-3-flash-preview_high","observed_competition_rank":30,"organization":"Google DeepMind","original":{"denominator":66,"numerator":59,"percent":"89.39%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":7,"clean":{"denominator":1323,"numerator":1213,"percent":"91.69%"},"clean_question_n":189,"configuration_date":"2026-02-06","configuration_id":"gpqa_cfg_v1_4cff4744741c24d1db4b97e733fde765e5f1c2056705a760369c260e9001e3d0","delta":{"denominator":29106,"numerator":919,"percentage_points":"+3.16 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_4a49b8bc4ef67cb52c4630ef9d37fa289e44c5df0580843e17425b64798c4739","evidence_class":"derived_uniform","evidence_variant_id":"gpqa_evd_v1_53194c97190cec47bf239eed235fa952bb1155e0d6a78b10ff7465d767b9ba83","model_configuration":"claude-opus-4-6_64K","observed_competition_rank":32,"organization":"Anthropic","original":{"denominator":462,"numerator":409,"percent":"88.53%"},"original_question_n":198,"protocol":"mean@7","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"derived_uniform_epoch_subset"},{"attempts_per_question":1,"clean":{"denominator":189,"numerator":173,"percent":"91.53%"},"clean_question_n":189,"configuration_date":"2026-08-06","configuration_id":"gpqa_cfg_v1_1752768af2b13dfd2f0a9df6802e29b14ab9387775bdd045c8ea1c437f8607f9","delta":{"denominator":4158,"numerator":131,"percentage_points":"+3.15 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_570b0a07b609fc7ca3c59be7f25e500bef27dce875b4a2e540fc1d41e286c20c","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_fa1e942e227b0d36cfcc50b1a9f58790f11c13327aa920e4cf1e6580d6cd1a34","model_configuration":"claude-opus-4-6_max","observed_competition_rank":33,"organization":"Anthropic","original":{"denominator":198,"numerator":175,"percent":"88.38%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":8,"clean":{"denominator":756,"numerator":691,"percent":"91.40%"},"clean_question_n":189,"configuration_date":"2025-12-11","configuration_id":"gpqa_cfg_v1_08b0ee7e2e105ac8e5ed96cd2f47a42dff3bba9cd0b1c886aac128321d37784f","delta":{"denominator":3024,"numerator":97,"percentage_points":"+3.21 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_178cd776ee64853cf54697ca6eae803398dbded170cc215ba350fbb0affce958","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_8325f704751ccd8d53ca9996d72cb19b61a6f9d27e9243b9b4a2a474023c2c2c","model_configuration":"gpt-5.2-2025-12-11_high","observed_competition_rank":34,"organization":"OpenAI","original":{"denominator":144,"numerator":127,"percent":"88.19%"},"original_question_n":198,"protocol":"mean@8","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":1,"clean":{"denominator":189,"numerator":172,"percent":"91.01%"},"clean_question_n":189,"configuration_date":"2026-08-06","configuration_id":"gpqa_cfg_v1_18c8f9728ba3bccb250dcce8c3a5d52ac4be8aae99733f291cecb08e35b283db","delta":{"denominator":2079,"numerator":65,"percentage_points":"+3.13 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_45aea34803bca828427c03f1a9fd3840200b233b1ee16c224e8d982f1c636674","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_9560f7cb9c978b6e8f919d7675abd559fa0708481f617c01f7e4dc2cf26536da","model_configuration":"claude-opus-5_low","observed_competition_rank":35,"organization":"Anthropic","original":{"denominator":33,"numerator":29,"percent":"87.88%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":1,"clean":{"denominator":189,"numerator":172,"percent":"91.01%"},"clean_question_n":189,"configuration_date":"2026-08-16","configuration_id":"gpqa_cfg_v1_538540e64da939fe97f8685591c8f9e03cd907c322ec906eea6256ce46fc72df","delta":{"denominator":2079,"numerator":65,"percentage_points":"+3.13 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_3d4e5a3f8219c5d1cdcd32800a4256c4732de5820b869d7f2124f79b702a64f8","evidence_class":"targeted_recovery","evidence_variant_id":"gpqa_evd_v1_fbdd43a5165165c2b670ee70b7ce930b5f7520a8cb06bb17327db3b8256b3bbe","model_configuration":"glm-5","observed_competition_rank":35,"organization":"Z.ai (Zhipu AI)","original":{"denominator":33,"numerator":29,"percent":"87.88%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"complete_with_epoch_targeted_transport_recovery"},{"attempts_per_question":1,"clean":{"denominator":189,"numerator":172,"percent":"91.01%"},"clean_question_n":189,"configuration_date":"2026-08-07","configuration_id":"gpqa_cfg_v1_6ad1e1d106980a4cd18c39c471f9ddbfbff5865f9045d01824a081231d140126","delta":{"denominator":4158,"numerator":109,"percentage_points":"+2.62 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_cb255829d8ee31b7dfe578019dec2f345f71f58b4e39ada5063aba69a50b7382","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_f4bf55465a437307ef0725d6ccd555d935204df37560205c5a3e6d7efd41e927","model_configuration":"qwen3.6-plus","observed_competition_rank":35,"organization":"Alibaba","original":{"denominator":198,"numerator":175,"percent":"88.38%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":1,"clean":{"denominator":189,"numerator":172,"percent":"91.01%"},"clean_question_n":189,"configuration_date":"2026-08-07","configuration_id":"gpqa_cfg_v1_f9fede76cf79ca860917f59fb5ebf3126982f366fe21b16ef0e22d4d13f4038a","delta":{"denominator":2079,"numerator":65,"percentage_points":"+3.13 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_d8361abeb8e6acd4751b11badb2b386a2411608213c72001ac4522da086c93f1","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_1b43335efd685be40f7926a1a9b21834dfe490d6f861040f1f46e757dae3ef25","model_configuration":"kimi-k2.7-code","observed_competition_rank":35,"organization":"Moonshot","original":{"denominator":33,"numerator":29,"percent":"87.88%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":1,"clean":{"denominator":189,"numerator":172,"percent":"91.01%"},"clean_question_n":189,"configuration_date":"2026-08-06","configuration_id":"gpqa_cfg_v1_fe902b8590238e8946b66aae41d7ee7406b4975a2264ad5dc8639eef08bc33ac","delta":{"denominator":4158,"numerator":109,"percentage_points":"+2.62 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_2a3998731aae032047bf6cdf10d6214e0452a01a333e66aa6a4563bf93377a96","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_155d6aa5896c8d17cff48bb97ca705e21a1f690a4fd87441cb4ace09eb37a96a","model_configuration":"claude-opus-4-8_low","observed_competition_rank":35,"organization":"Anthropic","original":{"denominator":198,"numerator":175,"percent":"88.38%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":8,"clean":{"denominator":1512,"numerator":1375,"percent":"90.94%"},"clean_question_n":189,"configuration_date":"2025-12-11","configuration_id":"gpqa_cfg_v1_2c3d5349208f65704ed39c3181dd248e34283f155f2c5b6d9238c3eebf0c47b3","delta":{"denominator":16632,"numerator":509,"percentage_points":"+3.06 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_71d59e96915ee1c9f1e065c8a90b2578853a6ba822896791a753d1cd666ace6d","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_8a451064c46ccfaf597b3db38e4a4e6a6ba9ef8e48e8606e2246b5d9d7929ab0","model_configuration":"gpt-5.2-2025-12-11_medium","observed_competition_rank":40,"organization":"OpenAI","original":{"denominator":33,"numerator":29,"percent":"87.88%"},"original_question_n":198,"protocol":"mean@8","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":8,"clean":{"denominator":21,"numerator":19,"percent":"90.48%"},"clean_question_n":189,"configuration_date":"2026-01-28","configuration_id":"gpqa_cfg_v1_25b6ff3864c66bd8fca985a9bbb8460bb4e531eb06f3d77a0f494f3b913eb510","delta":{"denominator":11088,"numerator":323,"percentage_points":"+2.91 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_88e1f487b6103967cbd712dc8da2e698590c83ea10f5e78720c5eb14f83de23c","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_482634699135b907fc082cde0565498864c3156e5ad7d1ba04c678c58830a443","model_configuration":"fireworks/kimi-k2p5","observed_competition_rank":41,"organization":"Moonshot","original":{"denominator":1584,"numerator":1387,"percent":"87.56%"},"original_question_n":198,"protocol":"mean@8","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":1,"clean":{"denominator":21,"numerator":19,"percent":"90.48%"},"clean_question_n":189,"configuration_date":"2026-08-07","configuration_id":"gpqa_cfg_v1_5c21ac281f3717fbc022623ee94cf1b26302c0a4bedd910769f1e8033717239a","delta":{"denominator":77,"numerator":2,"percentage_points":"+2.60 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_cd28725ecd192d79320824311013bd267aaac131969bbe81362007435ea9e3f5","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_ffdd72fff3ccfc6be88e91097c15ef87e0be38b4780e9f1ce68aae731188d9c0","model_configuration":"qwen3.7-plus","observed_competition_rank":41,"organization":"Alibaba","original":{"denominator":33,"numerator":29,"percent":"87.88%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":1,"clean":{"denominator":21,"numerator":19,"percent":"90.48%"},"clean_question_n":189,"configuration_date":"2026-08-07","configuration_id":"gpqa_cfg_v1_e3b6650e66d0e9441ad0ed99a22261238f2ac8b4e5ede02f20cde8b6b5dd76c7","delta":{"denominator":1386,"numerator":43,"percentage_points":"+3.10 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_ffc346943686ed230a2d0f91fde947bf8cd434dd5bbc80188a285678a84ba876","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_01baead65449a08b3c3cce21e89541808b31b6242f5a2e5463e9c35452333126","model_configuration":"gpt-5.6-terra_low","observed_competition_rank":41,"organization":"OpenAI","original":{"denominator":198,"numerator":173,"percent":"87.37%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":1,"clean":{"denominator":21,"numerator":19,"percent":"90.48%"},"clean_question_n":189,"configuration_date":"2026-08-07","configuration_id":"gpqa_cfg_v1_fefffa8ea64c405943e964126ee6c767f5126e9879cdf3c2989670b3db7d61ba","delta":{"denominator":1386,"numerator":43,"percentage_points":"+3.10 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_88001e90b5523fd6dac83e0c41e3b7e835d362af3bca441242d12c32d47900cb","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_95397797b4a12df58b6492f30b469f0caaf6693aa46c9ee2bb567093236faf38","model_configuration":"qwen3.6-max-preview","observed_competition_rank":41,"organization":"Alibaba","original":{"denominator":198,"numerator":173,"percent":"87.37%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":8,"clean":{"denominator":1512,"numerator":1367,"percent":"90.41%"},"clean_question_n":189,"configuration_date":"2025-11-13","configuration_id":"gpqa_cfg_v1_7f8e8f18de3290be2ccfe3701c2d2a4c8a6512f34dc46b7db05c5915a94d40e5","delta":{"denominator":16632,"numerator":463,"percentage_points":"+2.78 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_d0ba371b076810f91175ff17795e3a89abf2616e381bc55dff6bd330861faaf8","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_42dfda83b547f30bf90fc46faf232f79087f3a861423aae0c5efee1c7b6dc543","model_configuration":"gpt-5.1-2025-11-13_high","observed_competition_rank":45,"organization":"OpenAI","original":{"denominator":396,"numerator":347,"percent":"87.63%"},"original_question_n":198,"protocol":"mean@8","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":1,"clean":{"denominator":189,"numerator":170,"percent":"89.95%"},"clean_question_n":189,"configuration_date":"2026-08-06","configuration_id":"gpqa_cfg_v1_5deb548701e55fbdff283a419bd403f4d1d5b178a1239bc33ecc4508382d6b91","delta":{"denominator":4158,"numerator":149,"percentage_points":"+3.58 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_7abfe4b2b794eb665fef12fc1a1e137207e333bcaf6baca00b3dd270ef65275f","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_dfd90981850dbbd7283dcd1e760b706e7f423fb94a3ca31f10920df8e474d5d8","model_configuration":"gemini-3.6-flash_low","observed_competition_rank":46,"organization":"Google DeepMind","original":{"denominator":22,"numerator":19,"percent":"86.36%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":1,"clean":{"denominator":189,"numerator":170,"percent":"89.95%"},"clean_question_n":189,"configuration_date":"2026-08-07","configuration_id":"gpqa_cfg_v1_8db4d324508b283804b4f687b846605f27d9446a2eb2c449127b198717926f5a","delta":{"denominator":2079,"numerator":64,"percentage_points":"+3.08 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_dc613f0a761c2857feb508ffcd540d632d23ed7bbab5e6c1c2acff6df1cd66f7","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_184f378cffcfabf6ab4fa02c4adfd83638aee252429f0ba058653c6ecb3e41fb","model_configuration":"gpt-5.4-mini-2026-03-17_xhigh","observed_competition_rank":46,"organization":"OpenAI","original":{"denominator":99,"numerator":86,"percent":"86.87%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":1,"clean":{"denominator":189,"numerator":170,"percent":"89.95%"},"clean_question_n":189,"configuration_date":"2026-07-15","configuration_id":"gpqa_cfg_v1_a9fa8925e31e2718da34308ac52093f3a67b1df4c9d64b4dd42be93e3f710655","delta":{"denominator":4158,"numerator":149,"percentage_points":"+3.58 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_1d8842cd707e4be369af820729ad0b74fff0b1ffbd484d9486044b330154ca40","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_2367b1c3f49024155d12edadedb0d216eb5c09be7d49d6a5f5e2b971ac96d54d","model_configuration":"gemini-3.5-flash_minimal","observed_competition_rank":46,"organization":"Google","original":{"denominator":22,"numerator":19,"percent":"86.36%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":1,"clean":{"denominator":189,"numerator":170,"percent":"89.95%"},"clean_question_n":189,"configuration_date":"2026-08-17","configuration_id":"gpqa_cfg_v1_f657ce567632b0af5c3025bd73318b5a4796ffe43af7c00f6a98baea08eea2b2","delta":{"denominator":2079,"numerator":64,"percentage_points":"+3.08 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_e53fe107892393c2cca4ab8b9572a08fae423556b3acb3a07e56eb39fafaf584","evidence_class":"project_native","evidence_variant_id":"gpqa_evd_v1_a5dcd773dbe55fda5fc636b9f84c9bd0a59d39eaddee70cffcaad64103e05720","model_configuration":"Grok 4.3 (high)","observed_competition_rank":46,"organization":"xAI","original":{"denominator":99,"numerator":86,"percent":"86.87%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_all_normal"},{"attempts_per_question":8,"clean":{"denominator":126,"numerator":113,"percent":"89.68%"},"clean_question_n":189,"configuration_date":"2026-02-20","configuration_id":"gpqa_cfg_v1_cf7a0ccdcb575260739a0107fbf38e7b46265e7cca9bb7f750093cf0cbf5cc6c","delta":{"denominator":693,"numerator":16,"percentage_points":"+2.31 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_539787a00fa51afcec3ccb65ccff97d320fb4694a812942a0833e0e8eb153471","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_36660bc48bc7969c439dc1c3957e2acbc6c27c7402ae011671ba60bc49221447","model_configuration":"claude-sonnet-4-6_32K","observed_competition_rank":50,"organization":"Anthropic","original":{"denominator":198,"numerator":173,"percent":"87.37%"},"original_question_n":198,"protocol":"mean@8","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":8,"clean":{"denominator":756,"numerator":677,"percent":"89.55%"},"clean_question_n":189,"configuration_date":"2025-10-29","configuration_id":"gpqa_cfg_v1_4f3803305451e61c9246b82a829be2d36ae588a015920201efed023d91dabfaa","delta":{"denominator":33264,"numerator":1123,"percentage_points":"+3.38 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_d987f92c42189d8e928c6ec2a35ab3b520867c90143588803e8bbe84b7306481","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_d65c0c339d9fb28eb80cdd3c8012c18af1eaf202129b8a347da9486b187ac48b","model_configuration":"gpt-5-2025-08-07_high","observed_competition_rank":51,"organization":"OpenAI","original":{"denominator":528,"numerator":455,"percent":"86.17%"},"original_question_n":198,"protocol":"mean@8","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":1,"clean":{"denominator":189,"numerator":169,"percent":"89.42%"},"clean_question_n":189,"configuration_date":"2026-08-06","configuration_id":"gpqa_cfg_v1_114b61ba4b46f7fbd32882dba76813956cbdfdc2782903c93f510f22a3b1fc54","delta":{"denominator":4158,"numerator":127,"percentage_points":"+3.05 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_f8b55dfa9d69764aed33a4fa19c62e976ae46b2f92ece4e21abdbf78bbdfdcc0","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_b264bec2407ede2a026598c2260682c73d6cfc7412a6e7a9596aa89c6b092fd9","model_configuration":"claude-opus-4-7_max","observed_competition_rank":52,"organization":"Anthropic","original":{"denominator":22,"numerator":19,"percent":"86.36%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":1,"clean":{"denominator":189,"numerator":169,"percent":"89.42%"},"clean_question_n":189,"configuration_date":"2026-08-06","configuration_id":"gpqa_cfg_v1_9fdc200cc70fb609af6cd2ea5c6ec5818e339138000d351fc142b07d30c01378","delta":{"denominator":2079,"numerator":74,"percentage_points":"+3.56 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_c4c9571f3a66edb9d81d78b50b0833b458ea5c0cff5ffdd1fa5ecadbcf484b3b","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_4e869a342ecda24ebbf73895bbe205af9ad2a2ddd416b459ca09df248261173b","model_configuration":"claude-fable-5_max","observed_competition_rank":52,"organization":"Anthropic","original":{"denominator":99,"numerator":85,"percent":"85.86%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":1,"clean":{"denominator":189,"numerator":169,"percent":"89.42%"},"clean_question_n":189,"configuration_date":"2026-08-07","configuration_id":"gpqa_cfg_v1_a0a4273f3e80a8760824125f9fe0ff0835270ab6ca17ddf3d6af4f4400e5ecd9","delta":{"denominator":2079,"numerator":74,"percentage_points":"+3.56 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_505ce6fbf3263877df0b03176c3eb6e48ab75c5465bd8e93f26b1d3c1492602c","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_0d87f39869e0bcbe38244e446fc4e00bcd437aa68bae612afa74825b598e395c","model_configuration":"qwen3.5-397b-a17b","observed_competition_rank":52,"organization":"Alibaba","original":{"denominator":99,"numerator":85,"percent":"85.86%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":1,"clean":{"denominator":189,"numerator":169,"percent":"89.42%"},"clean_question_n":189,"configuration_date":"2026-08-07","configuration_id":"gpqa_cfg_v1_c1890a410e08444aa2e81859f26860a80431a4a5028c50bc7f3514cbc2e50a82","delta":{"denominator":4158,"numerator":127,"percentage_points":"+3.05 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_c0da6efa4e9f19cddbdea469776df805e3444c48cefd686086017bff4f5bc1af","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_9c8a0e08aa9dbc17aab6da7a789f24d40afda3b4ac40259a5f8cd7bb4a47d695","model_configuration":"qwen3.5-397b-a17b_none","observed_competition_rank":52,"organization":"Alibaba","original":{"denominator":22,"numerator":19,"percent":"86.36%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":1,"clean":{"denominator":9,"numerator":8,"percent":"88.89%"},"clean_question_n":189,"configuration_date":"2026-08-07","configuration_id":"gpqa_cfg_v1_d6f20b74ffa7e4b76d5c03ebd92e8ebb6053e6c47efc1d83f30e9db73129cb4f","delta":{"denominator":33,"numerator":1,"percentage_points":"+3.03 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_74681dd1b2c1294340edec610a41dad935c775a6e62717aa6c6b11d2320be9b5","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_9363b6a95d4c5080b049e9e51f01065d4e54cc46c7771bca9f19e02654853141","model_configuration":"qwen3.6-27b","observed_competition_rank":56,"organization":"Alibaba","original":{"denominator":99,"numerator":85,"percent":"85.86%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":8,"clean":{"denominator":756,"numerator":671,"percent":"88.76%"},"clean_question_n":189,"configuration_date":"2025-11-24","configuration_id":"gpqa_cfg_v1_cad9ce1ea155bb333c5ebd7a19fec521f9db3f4590d2bf1ec498004a3a8f0362","delta":{"denominator":33264,"numerator":901,"percentage_points":"+2.71 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_8f0f1f40c4ae6fb2b37d51816719e6564c712c79f04cb6c95299b61460522874","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_1b5cc11fffab55b07f6d9640a719846848a6c214ad4abd1a27a0364b071a2511","model_configuration":"claude-opus-4-5-20251101_32K","observed_competition_rank":57,"organization":"Anthropic","original":{"denominator":1584,"numerator":1363,"percent":"86.05%"},"original_question_n":198,"protocol":"mean@8","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":7,"clean":{"denominator":1323,"numerator":1174,"percent":"88.74%"},"clean_question_n":189,"configuration_date":"2025-08-07","configuration_id":"gpqa_cfg_v1_7c3ec07ae3ad1a5f35ccd1d892d17caa2ff48036699a386811c33885de02853a","delta":{"denominator":14553,"numerator":482,"percentage_points":"+3.31 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_4064f9f302dd20b904afc43b14493d0a8bb8a5fcf6febc63c3c6b4412c7c1118","evidence_class":"derived_uniform","evidence_variant_id":"gpqa_evd_v1_7d210d7ef8e66f427900f281667fcd9ca91b814e5ac9d1218fd88c7f8a6f44db","model_configuration":"gpt-5-2025-08-07_medium","observed_competition_rank":58,"organization":"OpenAI","original":{"denominator":693,"numerator":592,"percent":"85.43%"},"original_question_n":198,"protocol":"mean@7","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"derived_uniform_epoch_subset"},{"attempts_per_question":8,"clean":{"denominator":168,"numerator":149,"percent":"88.69%"},"clean_question_n":189,"configuration_date":"2025-11-25","configuration_id":"gpqa_cfg_v1_d7b58af1465d20d5223ea9f8a5e54b764667d1cb51518b107c4dbfe5c939e1fd","delta":{"denominator":2772,"numerator":89,"percentage_points":"+3.21 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_e24e64b1867767e63662a1a29ec36856570f8e7c17f2479d7d156f4ce0b5326a","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_1445ad1e03ba2ff428c21e82355430097a475172456c0f79f66a9b2219e595a4","model_configuration":"claude-opus-4-5-20251101_16K","observed_competition_rank":59,"organization":"Anthropic","original":{"denominator":792,"numerator":677,"percent":"85.48%"},"original_question_n":198,"protocol":"mean@8","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":1,"clean":{"denominator":189,"numerator":167,"percent":"88.36%"},"clean_question_n":189,"configuration_date":"2026-08-07","configuration_id":"gpqa_cfg_v1_a7d0e1353fd8be2dc557c8c94f0706588043b35fd3151e4c1a3ca31801ecb166","delta":{"denominator":2079,"numerator":52,"percentage_points":"+2.50 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_e0c490303a55d69319dec22bff959acb555b73644e6b2c388e9043bace59ab6d","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_9a537c87ec56513fbcdf9abf92d98029fa6ae035190d848d8e01ced5f6603ced","model_configuration":"gemini-3.6-flash_minimal","observed_competition_rank":60,"organization":"Google DeepMind","original":{"denominator":99,"numerator":85,"percent":"85.86%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":1,"clean":{"denominator":189,"numerator":167,"percent":"88.36%"},"clean_question_n":189,"configuration_date":"2026-08-07","configuration_id":"gpqa_cfg_v1_c413680e8fd715c2789e6cbb466901dd7cf9f221fe5a9a6f5aab61e02c530288","delta":{"denominator":2079,"numerator":73,"percentage_points":"+3.51 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_59d0bd8a33443ddf2b37cfc2c208107de7dea73e825054b67d1074f55f0a4a4e","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_6b961fd55ef7315eab24f3b486b42c7cec1b605f80a795374b8fa76e31439cf5","model_configuration":"qwen3.5-plus","observed_competition_rank":60,"organization":"Alibaba","original":{"denominator":33,"numerator":28,"percent":"84.85%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":1,"clean":{"denominator":189,"numerator":167,"percent":"88.36%"},"clean_question_n":189,"configuration_date":"2026-08-07","configuration_id":"gpqa_cfg_v1_d255d05df7d6452741ae49d1c8831a4ac1bdeb8644fc8d5d6b034ded7651c060","delta":{"denominator":2079,"numerator":73,"percentage_points":"+3.51 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_c6a9c4a5c08a066c77e81fe8228a1174a0376b9d8ffab4570806764d9d0063d0","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_3785c6853487ce53ec1e2d0239ab56dc955496b1dd1e9813324d4c628900c02e","model_configuration":"kimi-k3_low","observed_competition_rank":60,"organization":"Moonshot","original":{"denominator":33,"numerator":28,"percent":"84.85%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":8,"clean":{"denominator":504,"numerator":445,"percent":"88.29%"},"clean_question_n":189,"configuration_date":"2025-11-16","configuration_id":"gpqa_cfg_v1_457d4d9de46a09e8dda7707cbea4e03840a5686a817ef7dbd62b104c41ed8b9e","delta":{"denominator":1232,"numerator":37,"percentage_points":"+3.00 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_f24d6a067fc4d63a07a09e02dd0a7c8c5977b7e161958929a780337da209cc6d","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_99ceec1a96535e7c294d08862aad18ad46ed34e353d04fc1190c089cae51276c","model_configuration":"gemini-2.5-pro","observed_competition_rank":63,"organization":"Google DeepMind","original":{"denominator":1584,"numerator":1351,"percent":"85.29%"},"original_question_n":198,"protocol":"mean@8","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":1,"clean":{"denominator":189,"numerator":166,"percent":"87.83%"},"clean_question_n":189,"configuration_date":"2025-06-05","configuration_id":"gpqa_cfg_v1_2512d5e4965dc519e0cfe48a64b1f7f33961e8ebe6df554e0815a1b4c98f9bd3","delta":{"denominator":2079,"numerator":62,"percentage_points":"+2.98 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_b0f8ef5dae0e538e7eac0a1deed2e9f6384bc2db831c3b6b3a7f74c6158b0299","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_1c5b515b9968d1975e565bea3e248fed8c94496529cf7af2ac2988b0bf61e97f","model_configuration":"gemini-2.5-pro-preview-06-05","observed_competition_rank":64,"organization":"Google DeepMind","original":{"denominator":33,"numerator":28,"percent":"84.85%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":1,"clean":{"denominator":189,"numerator":166,"percent":"87.83%"},"clean_question_n":189,"configuration_date":"2026-08-06","configuration_id":"gpqa_cfg_v1_7422854d7e27c6743941108a0a9a0f278c02b92c202c12726abb5c9bc544adf5","delta":{"denominator":4158,"numerator":103,"percentage_points":"+2.48 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_007966d785ed20dd83217409daa7620a16ff82f40c2b1dd8da0f34590b53b297","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_39247dc711ff5ce6f23d19de68453a2aef23f1f158b08e99cf733dc15655f3f4","model_configuration":"claude-opus-4-8_none","observed_competition_rank":64,"organization":"Anthropic","original":{"denominator":198,"numerator":169,"percent":"85.35%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":1,"clean":{"denominator":189,"numerator":166,"percent":"87.83%"},"clean_question_n":189,"configuration_date":"2026-08-07","configuration_id":"gpqa_cfg_v1_c11ad0dd40116f32651cfd124cde890429cc3ff2cb1a8b1b529c9d58661f9850","delta":{"denominator":2079,"numerator":62,"percentage_points":"+2.98 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_6893083f0ef2fba44be6d48e4ab8e77f95e774692d113149cb4b0fdc5cc57a68","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_7b24b1da63ad62151bdc7d36a8f372843d4664e35829971354750ad1a70b5b94","model_configuration":"qwen3.6-27b_none","observed_competition_rank":64,"organization":"Alibaba","original":{"denominator":33,"numerator":28,"percent":"84.85%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":1,"clean":{"denominator":189,"numerator":166,"percent":"87.83%"},"clean_question_n":189,"configuration_date":"2026-07-15","configuration_id":"gpqa_cfg_v1_d03c3f1680526d4f61dc66ec0cfd3f39d637ecad3f65e348f985bb848fc3c8e7","delta":{"denominator":2079,"numerator":62,"percentage_points":"+2.98 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_2e9ebbffb2f4446f3d4354b423cda97a3f1b88a383fc75157db4feff80d8b54f","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_7e88c3c69bcc6af2eb0a12752dc66994a940cfbff475a81429f406c322692e3f","model_configuration":"gpt-5.4-2026-03-05_low","observed_competition_rank":64,"organization":"OpenAI","original":{"denominator":33,"numerator":28,"percent":"84.85%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":1,"clean":{"denominator":189,"numerator":166,"percent":"87.83%"},"clean_question_n":189,"configuration_date":"2026-08-07","configuration_id":"gpqa_cfg_v1_efb979697550c6b7b9d2f7fbf5e13c8cdcdc7f6b1ed66a691f8f08a9078593bc","delta":{"denominator":2079,"numerator":62,"percentage_points":"+2.98 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_8adf4c17b0dd41a26412938738bf1bc4c241411f3bfbf92cd828b502e3e20b1d","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_6b9efd62f14bc0f7a0b3a80a1eeaf7e87b621361f1dc52119f98cfbd275a0873","model_configuration":"qwen3.6-35b-a3b_none","observed_competition_rank":64,"organization":"Alibaba","original":{"denominator":33,"numerator":28,"percent":"84.85%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":8,"clean":{"denominator":1512,"numerator":1327,"percent":"87.76%"},"clean_question_n":189,"configuration_date":"2025-11-17","configuration_id":"gpqa_cfg_v1_0ec60fbcfbd0a5dc9ad3b1d16e08bf7c3678cacdd15194446fb742b017dc743b","delta":{"denominator":33264,"numerator":907,"percentage_points":"+2.73 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_8643462e411041d61c09b3c93ef9495747941781878ee9de507ccce76e9782d8","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_36f4bde2c0391d4f5f492f677c315435eb53db6d72b8a92e2d19b35ec273d51f","model_configuration":"gpt-5.1-2025-11-13_medium","observed_competition_rank":69,"organization":"OpenAI","original":{"denominator":528,"numerator":449,"percent":"85.04%"},"original_question_n":198,"protocol":"mean@8","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":1,"clean":{"denominator":63,"numerator":55,"percent":"87.30%"},"clean_question_n":189,"configuration_date":"2025-03-31","configuration_id":"gpqa_cfg_v1_c254d074bf7d4d562416e11363147be58eab58afb4c70c33aee01ccb278b50a8","delta":{"denominator":231,"numerator":8,"percentage_points":"+3.46 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_c4291c1149beaaeedd87a1e987d06d3384f9ba33ddba4e80416ddc38771a396d","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_8887a5d926c0f619f97426e65b45cf84d7ecc002ecc510f8a9df014e9bdd355b","model_configuration":"gemini-2.5-pro-exp-03-25","observed_competition_rank":70,"organization":"Google DeepMind","original":{"denominator":99,"numerator":83,"percent":"83.84%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":1,"clean":{"denominator":189,"numerator":164,"percent":"86.77%"},"clean_question_n":189,"configuration_date":"2026-08-07","configuration_id":"gpqa_cfg_v1_1df69af1ab0ce9fcc3c822f72011312d0cb297c6ba8c857f0c29740914d904fd","delta":{"denominator":2079,"numerator":61,"percentage_points":"+2.93 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_814a3fcb9d9f82fcc6444bd32c8d97992010cbcaf704280c1f4b682f692cb48a","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_40b347dd19f8820810d1e9ab840304435b17392d485ef2f88a5dbc93537370cb","model_configuration":"qwen3.6-35b-a3b","observed_competition_rank":71,"organization":"Alibaba","original":{"denominator":99,"numerator":83,"percent":"83.84%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":1,"clean":{"denominator":189,"numerator":164,"percent":"86.77%"},"clean_question_n":189,"configuration_date":"2026-08-06","configuration_id":"gpqa_cfg_v1_9b2222a171dd46e01575f8e6e673c3b8b1659464ddb17ad4146c5e1da3a26a81","delta":{"denominator":378,"numerator":13,"percentage_points":"+3.44 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_ac9984b33f79389262288d99038c56128b471ee5a1e2476e05b98dd232e3d21a","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_ffff552e9fd10ae3b515b91b12b881f737cf27f85b92163d1d0a7143e73a9bc0","model_configuration":"claude-fable-5_high","observed_competition_rank":71,"organization":"Anthropic","original":{"denominator":6,"numerator":5,"percent":"83.33%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":1,"clean":{"denominator":189,"numerator":164,"percent":"86.77%"},"clean_question_n":189,"configuration_date":"2026-07-13","configuration_id":"gpqa_cfg_v1_eb8810fec6811f5a48d750f48d865427a72cac5e58054882006e251bfa0c646c","delta":{"denominator":378,"numerator":13,"percentage_points":"+3.44 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_2353d94ce816c022abcc725ec98763afab9beb630bafdf4d8f4b53929c8b82a9","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_1c402b295c3e04ffba4e8f41d2514d5dd0a8c690e48240a464ee9e1f92943521","model_configuration":"claude-sonnet-4-6_high","observed_competition_rank":71,"organization":"Anthropic","original":{"denominator":6,"numerator":5,"percent":"83.33%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":8,"clean":{"denominator":189,"numerator":164,"percent":"86.77%"},"clean_question_n":189,"configuration_date":"2026-04-15","configuration_id":"gpqa_cfg_v1_fee207fd49a1b7d70751742c0f126d3d087b989d89849747726ee9468fdf26f8","delta":{"denominator":8316,"numerator":265,"percentage_points":"+3.19 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_0e4b6e4c02d3232be69a1c3170e5e8ff805e13c1eb71d62ced537c556c817201","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_f939de26b6886fbb10f4be0e85613b485726fae0db2aa4d5de86f1a5db887a63","model_configuration":"gpt-5.4-mini-2026-03-17_high","observed_competition_rank":71,"organization":"OpenAI","original":{"denominator":396,"numerator":331,"percent":"83.59%"},"original_question_n":198,"protocol":"mean@8","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":8,"clean":{"denominator":126,"numerator":109,"percent":"86.51%"},"clean_question_n":189,"configuration_date":"2025-11-11","configuration_id":"gpqa_cfg_v1_f76c85c895ef2ab274d28dcefa237ebea5b991f40fdb01ee93070b5634a7ca73","delta":{"denominator":5544,"numerator":127,"percentage_points":"+2.29 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_11891105ed0f8462d7e8df193d42d17c0293f0ff9eb109ebb09098dc4d3afe7b","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_77b79be9e51b33386fd343cef692dea70e6249188126a76ed404f6bbd83c0b6c","model_configuration":"kimi-k2-thinking-turbo","observed_competition_rank":75,"organization":"Moonshot","original":{"denominator":792,"numerator":667,"percent":"84.22%"},"original_question_n":198,"protocol":"mean@8","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":1,"clean":{"denominator":189,"numerator":163,"percent":"86.24%"},"clean_question_n":189,"configuration_date":"2026-07-13","configuration_id":"gpqa_cfg_v1_5e269fbf7b862f78bdfb58b4e2d6a616de3f24a13172a171b7911c19e5882ee3","delta":{"denominator":378,"numerator":11,"percentage_points":"+2.91 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_9f54b987dc70091682cf9a98d5d242268c0a722fcbce4ad077b2c539c710abb7","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_57cbfb79bddc645ebc4a73a8fe89cafca7357cdaf497dfddd69160ac963e4318","model_configuration":"claude-sonnet-4-6_medium","observed_competition_rank":76,"organization":"Anthropic","original":{"denominator":6,"numerator":5,"percent":"83.33%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":1,"clean":{"denominator":189,"numerator":163,"percent":"86.24%"},"clean_question_n":189,"configuration_date":"2026-08-07","configuration_id":"gpqa_cfg_v1_ed668418bc3a7abc717016181a36d45c94836c4061801b735c989f31101fc953","delta":{"denominator":2079,"numerator":71,"percentage_points":"+3.42 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_1c9e192f85b264d68234f5d769b6214dc0efe4da9362a5329f0bd6e5ff330295","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_09e49c686372b8cb2f26f51a86830f6fa6da26694e7a114a4f24bce18f7e4848","model_configuration":"gpt-5.6-sol_none","observed_competition_rank":76,"organization":"OpenAI","original":{"denominator":99,"numerator":82,"percent":"82.83%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":7,"clean":{"denominator":441,"numerator":380,"percent":"86.17%"},"clean_question_n":189,"configuration_date":"2025-12-16","configuration_id":"gpqa_cfg_v1_b79f40ee05c16a27a3fae7d506fb2e654418d9d40244460a93ad4d4fc6b45468","delta":{"denominator":882,"numerator":25,"percentage_points":"+2.83 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_3ced4fbc7a914e379f2ddb1f5d74538c090671f9fd3ed597cc9004240a7b4384","evidence_class":"derived_uniform","evidence_variant_id":"gpqa_evd_v1_6de8ed01eeaadd5ec84d2baa19feac6cc625ae736e7979d3469c085b7137091c","model_configuration":"deepseek-reasoner","observed_competition_rank":78,"organization":"DeepSeek","original":{"denominator":6,"numerator":5,"percent":"83.33%"},"original_question_n":198,"protocol":"mean@7","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"derived_uniform_epoch_subset"},{"attempts_per_question":8,"clean":{"denominator":504,"numerator":433,"percent":"85.91%"},"clean_question_n":189,"configuration_date":"2025-12-17","configuration_id":"gpqa_cfg_v1_30fd0582710efe31443fff4d850caba23883e98a0a3df61e9dcbb85e5fb3ad3e","delta":{"denominator":924,"numerator":25,"percentage_points":"+2.71 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_1c256229a72afdf84b79498fc65ea39b9a088b233fc9fdd90ee8fb93496d40e8","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_5982441681aa590d4c027f2ae0243505f6c56f206f82a611e9efe4c9202c1614","model_configuration":"gemini-3-flash-preview","observed_competition_rank":79,"organization":"Google DeepMind","original":{"denominator":792,"numerator":659,"percent":"83.21%"},"original_question_n":198,"protocol":"mean@8","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":1,"clean":{"denominator":7,"numerator":6,"percent":"85.71%"},"clean_question_n":189,"configuration_date":"2025-10-28","configuration_id":"gpqa_cfg_v1_0bdf7182d5f04c31013cbf0810c5d35c04d75be94de6c6030b1439ce8d8f0278","delta":{"denominator":1386,"numerator":47,"percentage_points":"+3.39 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_88e47969d46da216fc880a0887465e0ae54b02d16fbd32933b18dd75f3d8ab1a","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_f51c42739dfd105b28ddb7ab432843e2f89aabfec69eda041292bdd0a1e2af8a","model_configuration":"claude-sonnet-4-5-20250929_59K","observed_competition_rank":80,"organization":"Anthropic","original":{"denominator":198,"numerator":163,"percent":"82.32%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":1,"clean":{"denominator":7,"numerator":6,"percent":"85.71%"},"clean_question_n":189,"configuration_date":"2026-08-07","configuration_id":"gpqa_cfg_v1_73606fc2f28e306f4091aca42290cd5715fed6ba0d38ff358cc6a8ec8c2cc294","delta":{"denominator":1386,"numerator":47,"percentage_points":"+3.39 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_ecfbe5ae897e0b755a15fc6c08ccdf595d1570019e11de6a5441cde295325a78","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_dac8aeaa9e559eae6dec72c411c74f09e7d7cc4345802beffe1c711904671881","model_configuration":"qwen3.5-flash","observed_competition_rank":80,"organization":"Alibaba","original":{"denominator":198,"numerator":163,"percent":"82.32%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":1,"clean":{"denominator":7,"numerator":6,"percent":"85.71%"},"clean_question_n":189,"configuration_date":"2026-08-07","configuration_id":"gpqa_cfg_v1_91d52572fee638e3cdcba8d3d9db001f9c9ce30df32420006ed56c7f4c01fca1","delta":{"denominator":42,"numerator":1,"percentage_points":"+2.38 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_171cf451c16bc9a1bf1bc32e261cf4616708b5f4be30c11a4fd6733b290ab59d","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_76b04aaf9ed0e8d643348ea46e48e52224ed5df3fd7fe47410106a956c73ce5d","model_configuration":"qwen3.6-flash","observed_competition_rank":80,"organization":"Alibaba","original":{"denominator":6,"numerator":5,"percent":"83.33%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":1,"clean":{"denominator":7,"numerator":6,"percent":"85.71%"},"clean_question_n":189,"configuration_date":"2026-08-06","configuration_id":"gpqa_cfg_v1_e4fc3d3f5f663e3176f37cb7a7a9414683ad660a66fd87026e1ba802553ac32a","delta":{"denominator":42,"numerator":1,"percentage_points":"+2.38 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_afd4e1102acfce0a7be890d674343db5836d1bd5e4dac9122614973fbf39a7a0","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_3a596d2e4295830d1c2ac6fa59311afe7cd4ade34ea5e652f41fa63a15c9ca5f","model_configuration":"gemini-3.5-flash-lite_high","observed_competition_rank":80,"organization":"Google DeepMind","original":{"denominator":6,"numerator":5,"percent":"83.33%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":8,"clean":{"denominator":216,"numerator":185,"percent":"85.65%"},"clean_question_n":189,"configuration_date":"2025-12-11","configuration_id":"gpqa_cfg_v1_aa564e65d8869f7ebd98b3ce39ff8e2e4e55cefa67b9e71802509a553137b9b3","delta":{"denominator":1188,"numerator":35,"percentage_points":"+2.95 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_10801929840f40ed71000744d2c8be3015711ba3e564ea85b8dcd2e6cc8698d6","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_0b5102859a1d1695bf2db4b37b7c23c96ea393213bfd8639459f5bd44de2d768","model_configuration":"gpt-5.2-2025-12-11_low","observed_competition_rank":84,"organization":"OpenAI","original":{"denominator":792,"numerator":655,"percent":"82.70%"},"original_question_n":198,"protocol":"mean@8","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":1,"clean":{"denominator":27,"numerator":23,"percent":"85.19%"},"clean_question_n":189,"configuration_date":"2026-08-07","configuration_id":"gpqa_cfg_v1_636869f8e677b100d9390e305daa004e4d3cb6ad815d50d572ee8f9ea78baf1f","delta":{"denominator":594,"numerator":17,"percentage_points":"+2.86 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_f5a5438f7b8c161627cb8de0b374ccbefa3ba38c75404f423d0fadf85bfc5f2d","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_4a61556655d70c802cbc91fb6fc7184525bb823a524da04da3736a5fd36b9f30","model_configuration":"gpt-5.6-luna_low","observed_competition_rank":85,"organization":"OpenAI","original":{"denominator":198,"numerator":163,"percent":"82.32%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":1,"clean":{"denominator":27,"numerator":23,"percent":"85.19%"},"clean_question_n":189,"configuration_date":"2026-08-07","configuration_id":"gpqa_cfg_v1_e7c76115b41d855efcf0f7ae19b2b6d7a23a44cfa36c8e7a39886451c18c0663","delta":{"denominator":594,"numerator":17,"percentage_points":"+2.86 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_e6638e326bb3918bfd4455ca78a1f45c7ce4c0bff2025c897f8537e38257012b","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_203ab8170562b95e7b03b6991d2986c568fe66f69f7e984309afe47250368097","model_configuration":"qwen3.7-flash","observed_competition_rank":85,"organization":"Alibaba","original":{"denominator":198,"numerator":163,"percent":"82.32%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":8,"clean":{"denominator":756,"numerator":643,"percent":"85.05%"},"clean_question_n":189,"configuration_date":"2025-04-16","configuration_id":"gpqa_cfg_v1_9eb89da0bad3d09e7ddb8a304b1163ece8e34a5dc839f463fab250c970819ca6","delta":{"denominator":8316,"numerator":269,"percentage_points":"+3.23 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_d0173bf897b84effeca3bbf845afe3773e13a3ebe3a577039d76cf9a084a67bd","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_672bab37c4befe1323a7c63e129290e70301302a02097a114b31eefe261601aa","model_configuration":"o3-2025-04-16_high","observed_competition_rank":87,"organization":"OpenAI","original":{"denominator":11,"numerator":9,"percent":"81.82%"},"original_question_n":198,"protocol":"mean@8","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":1,"clean":{"denominator":189,"numerator":160,"percent":"84.66%"},"clean_question_n":189,"configuration_date":"2026-08-16","configuration_id":"gpqa_cfg_v1_028330d0cbeadddca05767705c635f8f34aa6f91322bce88dd9bd97968f4a05a","delta":{"denominator":2079,"numerator":59,"percentage_points":"+2.84 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_39f293bd3e07a520963865e115317a17f07e383f5d5743a1e7d98917d7cfb91d","evidence_class":"targeted_recovery","evidence_variant_id":"gpqa_evd_v1_dfca09330d2c1339c6a87d2d4f2f9a6a498a98ef89616bbe63fbeece5147e5b4","model_configuration":"claude-sonnet-4-5-20250929_32K","observed_competition_rank":88,"organization":"Anthropic","original":{"denominator":11,"numerator":9,"percent":"81.82%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"complete_with_epoch_targeted_transport_recovery"},{"attempts_per_question":1,"clean":{"denominator":189,"numerator":160,"percent":"84.66%"},"clean_question_n":189,"configuration_date":"2026-08-06","configuration_id":"gpqa_cfg_v1_5d0b6fa9beb00bd723082609126a60e2124f2f089224e8935b7733ee6207c2d7","delta":{"denominator":2079,"numerator":59,"percentage_points":"+2.84 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_5515cdbfedd9d940487b09a63e43e466478b0beb2320b9e0f21fc42720664d6d","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_391cc973ca75513e6dc11c93ac32a504b912b4a2b07bb145ed77f3bb6584a8f7","model_configuration":"gemini-3.1-flash-lite_high","observed_competition_rank":88,"organization":"Google","original":{"denominator":11,"numerator":9,"percent":"81.82%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":1,"clean":{"denominator":63,"numerator":53,"percent":"84.13%"},"clean_question_n":189,"configuration_date":"2026-08-07","configuration_id":"gpqa_cfg_v1_32831c12727d118cdd4d29d933e064c6f5efeb1e888228cb3ccaef5ca5e5a6c5","delta":{"denominator":693,"numerator":23,"percentage_points":"+3.32 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_dcb21a23564f01cdc1c264ade404b71aa500e53a3e399d757f8c497308719619","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_18d8b74684579952ebe9cf782aff1b9a24f464f0d5095291c2fad65ab2642db6","model_configuration":"qwen3.7-flash_none","observed_competition_rank":90,"organization":"Alibaba","original":{"denominator":99,"numerator":80,"percent":"80.81%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":1,"clean":{"denominator":63,"numerator":53,"percent":"84.13%"},"clean_question_n":189,"configuration_date":"2026-08-07","configuration_id":"gpqa_cfg_v1_8f361ae1265b1142ff8c109e0b1829cdd625d41f3ee9506a6cdcdbab3641ee73","delta":{"denominator":693,"numerator":16,"percentage_points":"+2.31 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_88dda16a219b4deefb03b1b8c48485c39684eba5099d7bb57271704b446d2532","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_665b4a95100cff19984ec1fe10ff9d2230434d77d3c7a09e39ee3e3cbb8a4ff7","model_configuration":"qwen3.7-plus_none","observed_competition_rank":90,"organization":"Alibaba","original":{"denominator":11,"numerator":9,"percent":"81.82%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":1,"clean":{"denominator":189,"numerator":158,"percent":"83.60%"},"clean_question_n":189,"configuration_date":"2026-08-06","configuration_id":"gpqa_cfg_v1_46e58f3c8d85f65c9905d98e961570ffe8b923db2111db25bcf593e5a586ba5b","delta":{"denominator":4158,"numerator":137,"percentage_points":"+3.29 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_203f5558a45d7350437aae82e52a8cd851089e49ce2a62abe6f2a1598ad7df4a","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_b70737936b1bfc0ef3b3501ad10683d41623c62dc2fb664c7eef286ecbd01c26","model_configuration":"claude-sonnet-5_max","observed_competition_rank":92,"organization":"Anthropic","original":{"denominator":66,"numerator":53,"percent":"80.30%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":1,"clean":{"denominator":189,"numerator":158,"percent":"83.60%"},"clean_question_n":189,"configuration_date":"2026-08-07","configuration_id":"gpqa_cfg_v1_8ce85e5a084acc7fe3dcf3183d33f605314c4189cf5aec161b8576be34f00bd1","delta":{"denominator":2079,"numerator":58,"percentage_points":"+2.79 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_dfc5597ff3f4619ab2c684c61d8977155d5597fbcb6df734467f38c9f4af5c60","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_a4dbad6bcf403b3842bdf7f7d57b60ba54618170886586f899fc4cc7beab88c3","model_configuration":"o3-2025-04-16_medium","observed_competition_rank":92,"organization":"OpenAI","original":{"denominator":99,"numerator":80,"percent":"80.81%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":1,"clean":{"denominator":189,"numerator":157,"percent":"83.07%"},"clean_question_n":189,"configuration_date":"2026-01-29","configuration_id":"gpqa_cfg_v1_8481fd6a05a2aa64f8278e43afe1cb7914779e0b788942a039de1af847ce744b","delta":{"denominator":4158,"numerator":115,"percentage_points":"+2.77 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_bdbf986b2d3259bd5c00ee776060fe30b54bacb6b4aff74149a800bf87be7cb9","evidence_class":"derived_uniform","evidence_variant_id":"gpqa_evd_v1_0fd65c0260532a12b47d3506b0ab22b48168dba2e1e02c07a0a9deb2ec82de1f","model_configuration":"glm-4.7","observed_competition_rank":94,"organization":"Z.ai (Zhipu AI)","original":{"denominator":66,"numerator":53,"percent":"80.30%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"derived_uniform_epoch_subset"},{"attempts_per_question":8,"clean":{"denominator":252,"numerator":209,"percent":"82.94%"},"clean_question_n":189,"configuration_date":"2025-04-16","configuration_id":"gpqa_cfg_v1_b8ff63f0f960b4e2f35e0a38f75a361a14b4ce8d403568ee967f98dc1ff60ed7","delta":{"denominator":1232,"numerator":41,"percentage_points":"+3.33 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_9d512d9f0cf52e51fefcc86a1269f13b4e27fffc906225428130f2ff13d7a1ec","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_2ab2a1cf44a4b3cb002d148f563c852e367cd00c47aa6fed450290047551b0b1","model_configuration":"o4-mini-2025-04-16_high","observed_competition_rank":95,"organization":"OpenAI","original":{"denominator":1584,"numerator":1261,"percent":"79.61%"},"original_question_n":198,"protocol":"mean@8","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":8,"clean":{"denominator":168,"numerator":139,"percent":"82.74%"},"clean_question_n":189,"configuration_date":"2025-11-24","configuration_id":"gpqa_cfg_v1_9d62b338e275defb1013265362d0e2e54512238ef149f471ef7b2873e01a0b0d","delta":{"denominator":924,"numerator":19,"percentage_points":"+2.06 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_35ca9ef5130ca50321fbfd90738cb342d545fa463d231948898b65ec476cab64","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_d5547fadfac67fbb8f45b56678573e2f1a140b03887f0cf581f3cc0228953155","model_configuration":"claude-opus-4-5-20251101","observed_competition_rank":96,"organization":"Anthropic","original":{"denominator":88,"numerator":71,"percent":"80.68%"},"original_question_n":198,"protocol":"mean@8","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":1,"clean":{"denominator":63,"numerator":52,"percent":"82.54%"},"clean_question_n":189,"configuration_date":"2026-07-15","configuration_id":"gpqa_cfg_v1_1ebdf14944b16525b20d369d7989c36385e707ef96468f681ac093d3ef05807e","delta":{"denominator":693,"numerator":19,"percentage_points":"+2.74 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_75401b34ec4d25283ce1395b7ece31ff4524cd468cdeba3429407a8da6e8cd8a","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_078a842bcb3e1914b31eb65db36bff3347e6d5fd2e782038b8737f54e5506c57","model_configuration":"o3-2025-04-16_low","observed_competition_rank":97,"organization":"OpenAI","original":{"denominator":99,"numerator":79,"percent":"79.80%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":1,"clean":{"denominator":63,"numerator":52,"percent":"82.54%"},"clean_question_n":189,"configuration_date":"2025-10-28","configuration_id":"gpqa_cfg_v1_7855e86e24fde2ea81d498e4c025c7acf2b0fe6d817cfe394ef7306ce637d247","delta":{"denominator":693,"numerator":26,"percentage_points":"+3.75 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_3ba386640515238cce493e6e6214e420a8059589db6fb0812ded064d8cc542bc","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_e191866acc8afba0d0bfd1e173f45f1af06507777c8d67c91ad414558d64d708","model_configuration":"claude-sonnet-4-5-20250929_16K","observed_competition_rank":97,"organization":"Anthropic","original":{"denominator":33,"numerator":26,"percent":"78.79%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":2,"clean":{"denominator":378,"numerator":311,"percent":"82.28%"},"clean_question_n":189,"configuration_date":"2025-12-11","configuration_id":"gpqa_cfg_v1_16b158ba0a71b9966a8c0e86927eb1da15d0ee18a198e905d951edda2884f53d","delta":{"denominator":8316,"numerator":185,"percentage_points":"+2.22 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_4ac367431f74dbfcd34e9f8b7882af29e08138f13020d3729e1d3fae75aa95a5","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_805e3318ae823eaf6a2c7bfeaae0d35cc1194cff3f81ab3b1a84975eb384cc33","model_configuration":"Qwen3-235B-A22B-Thinking-2507","observed_competition_rank":99,"organization":"Alibaba","original":{"denominator":396,"numerator":317,"percent":"80.05%"},"original_question_n":198,"protocol":"mean@2","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":1,"clean":{"denominator":189,"numerator":155,"percent":"82.01%"},"clean_question_n":189,"configuration_date":"2026-08-06","configuration_id":"gpqa_cfg_v1_0780c6b4ba641120d8293909ee3b41d284184448300fe82845ccd82ed21a22a8","delta":{"denominator":2079,"numerator":67,"percentage_points":"+3.22 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_47a174a410566d1d53939cd53d53bb0a9d35d6e888f166d50d2f8be2551da5ec","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_7e4b216cd43665ee130c572b462caad73d67ffa01fa6214de04878c9e9f3ff52","model_configuration":"claude-fable-5_low","observed_competition_rank":100,"organization":"Anthropic","original":{"denominator":33,"numerator":26,"percent":"78.79%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":1,"clean":{"denominator":27,"numerator":22,"percent":"81.48%"},"clean_question_n":189,"configuration_date":"2026-08-06","configuration_id":"gpqa_cfg_v1_72074cf78c83a95bdcd075539dbc4c33faf276be230e0adcaf85fd54c8a051ed","delta":{"denominator":297,"numerator":8,"percentage_points":"+2.69 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_2343c3fb0c777d1fae499e1d557cb8ed7cb87826dc5191bd8f6a21c53208ffee","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_c5ec14d09aa3a7b4aca91e27c2c73d542f79b32fe4b7c3db3150fe93a8c51d42","model_configuration":"claude-sonnet-4-6_max","observed_competition_rank":101,"organization":"Anthropic","original":{"denominator":33,"numerator":26,"percent":"78.79%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":8,"clean":{"denominator":216,"numerator":175,"percent":"81.02%"},"clean_question_n":189,"configuration_date":"2026-04-14","configuration_id":"gpqa_cfg_v1_850dd081ff0af5e05ae06d189189973570eabc763dcdf7b9c49aeb6b97112d76","delta":{"denominator":432,"numerator":11,"percentage_points":"+2.55 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_f0268b3f6b184e0701abb740ed28efa27bd588b9292ad027ebeacd5787b305f4","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_c93b6e72660dcbb86bcca25a2fe0a3780b2f0711ceefa8edb838cd88fe811031","model_configuration":"gpt-5.4-nano-2026-03-17_high","observed_competition_rank":102,"organization":"OpenAI","original":{"denominator":144,"numerator":113,"percent":"78.47%"},"original_question_n":198,"protocol":"mean@8","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":1,"clean":{"denominator":189,"numerator":152,"percent":"80.42%"},"clean_question_n":189,"configuration_date":"2025-05-22","configuration_id":"gpqa_cfg_v1_24e612fff1c0d93cfbb8e6ac4c6cd800e066c2b55d58c67b0615a42278896a49","delta":{"denominator":4158,"numerator":89,"percentage_points":"+2.14 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_ef8f65ab4b75285fd76a2d42d67149ff08799bcd26e49c56b343c8354f4a32e0","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_e842a50332b2f1acbf3c9ef5d7c9b255e8f2d0481ad5ca5137145ce0d9e4c36b","model_configuration":"claude-sonnet-4-20250514_32K","observed_competition_rank":103,"organization":"Anthropic","original":{"denominator":198,"numerator":155,"percent":"78.28%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":1,"clean":{"denominator":189,"numerator":152,"percent":"80.42%"},"clean_question_n":189,"configuration_date":"2025-08-05","configuration_id":"gpqa_cfg_v1_67756590ba7c7aa34e9e5076f7e90d0b4251a570d95d5dc77fcb42a6b854f2ee","delta":{"denominator":4158,"numerator":131,"percentage_points":"+3.15 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_a694f7431f860f256c8ff49ce93b03646e6ec24bbc0e969087c3ccf44a5ddd98","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_cfab3223935926a78a54266517b83c9ac0cf667c36e40b727410e634cd0c77b9","model_configuration":"claude-opus-4-1-20250805_16K","observed_competition_rank":103,"organization":"Anthropic","original":{"denominator":22,"numerator":17,"percent":"77.27%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":1,"clean":{"denominator":189,"numerator":152,"percent":"80.42%"},"clean_question_n":189,"configuration_date":"2026-08-07","configuration_id":"gpqa_cfg_v1_cc64a12d6bdca099340cd29a26afe83aeffe666f68bc05604bf7f4e7668632cf","delta":{"denominator":189,"numerator":5,"percentage_points":"+2.65 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_81d603eb7852d0c2381cccf6581f79b07c1a4587d624c00276deac0263a5404a","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_c1d607e0d4fc89e9b91829beebbcad48937243ac67b3be6691b1ad2d8b468d05","model_configuration":"o4-mini-2025-04-16_medium","observed_competition_rank":103,"organization":"OpenAI","original":{"denominator":9,"numerator":7,"percent":"77.78%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":1,"clean":{"denominator":189,"numerator":151,"percent":"79.89%"},"clean_question_n":189,"configuration_date":"2025-02-26","configuration_id":"gpqa_cfg_v1_582401c4d6cff8722feac8c0b9c6982da4d7180861f4671dfc33dd6a447680d5","delta":{"denominator":2079,"numerator":65,"percentage_points":"+3.13 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_9fbf8985c5ca2f5198c04f4e81c7b70e8964a76522f3abd5cbe7ca7da5d89927","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_8e4c9d6bc023349b56f42fd7442941038c1397f038dc1e688b19e7bf00982974","model_configuration":"claude-3-7-sonnet-20250219_16K","observed_competition_rank":106,"organization":"Anthropic","original":{"denominator":99,"numerator":76,"percent":"76.77%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":1,"clean":{"denominator":189,"numerator":151,"percent":"79.89%"},"clean_question_n":189,"configuration_date":"2025-03-10","configuration_id":"gpqa_cfg_v1_84be1c2e599b07f80232212023c305f33061879967f9ca1cd6e6a2aa92d3430f","delta":{"denominator":2079,"numerator":65,"percentage_points":"+3.13 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_46f3f78481720a365d57af638878d0ec26af8758ab8886c8eb51449c568cbe01","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_2ddbb950ca1860d699684d4213cdacf2330dcf95cbf97b94493b14e198e477dd","model_configuration":"claude-3-7-sonnet-20250219_32K","observed_competition_rank":106,"organization":"Anthropic","original":{"denominator":99,"numerator":76,"percent":"76.77%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":1,"clean":{"denominator":189,"numerator":151,"percent":"79.89%"},"clean_question_n":189,"configuration_date":"2025-08-05","configuration_id":"gpqa_cfg_v1_9e05b4b262c02d63871fd36e5097241f13ddb790d81aed45c3cdaa5f0066fe16","delta":{"denominator":2079,"numerator":65,"percentage_points":"+3.13 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_fedafd74cf86b8ed517d88a3f649bef8521dcb14017cc9efd5531eb98a06b277","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_7935ec345c8c47d45a8fc1278e7c2b591bfbc3ef956b43da9fc62ebd321319fb","model_configuration":"claude-opus-4-1-20250805_27K","observed_competition_rank":106,"organization":"Anthropic","original":{"denominator":99,"numerator":76,"percent":"76.77%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":1,"clean":{"denominator":189,"numerator":151,"percent":"79.89%"},"clean_question_n":189,"configuration_date":"2026-08-07","configuration_id":"gpqa_cfg_v1_bdda016ac92da795727c89d7eb197d32b9654ce78d9d202bf7a7aae394898d1b","delta":{"denominator":4158,"numerator":109,"percentage_points":"+2.62 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_441ffd9ecb95e12894b5ccd9cf32b7c3bb475f7721657ba3739f78bd1e1b3a9c","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_2294e88cba5d4836f3b78fa9fb13e3fd9138c00ce5d2e64a8f0b9f270bb58069","model_configuration":"gpt-5.5_none","observed_competition_rank":106,"organization":"OpenAI","original":{"denominator":22,"numerator":17,"percent":"77.27%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":8,"clean":{"denominator":504,"numerator":401,"percent":"79.56%"},"clean_question_n":189,"configuration_date":"2025-02-13","configuration_id":"gpqa_cfg_v1_5b98063364ead570d569428619f6b536cda9944d79c01cd3d1325c9b59abb6bb","delta":{"denominator":1848,"numerator":47,"percentage_points":"+2.54 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_6e14e47b5b7b1a91e8f040222b2da203dc2c7cd6d19251e70929a30fefa0e2db","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_d1dfeb319d5426dc160b9dea8728578465aa8b7a881e50f991c36d232ae4fcae","model_configuration":"o3-mini-2025-01-31_high","observed_competition_rank":110,"organization":"OpenAI","original":{"denominator":396,"numerator":305,"percent":"77.02%"},"original_question_n":198,"protocol":"mean@8","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":1,"clean":{"denominator":63,"numerator":50,"percent":"79.37%"},"clean_question_n":189,"configuration_date":"2025-02-13","configuration_id":"gpqa_cfg_v1_211e3b3ef5c20703200d0b2a23fabe754e645ec0860ac579ec797da9ffd88edb","delta":{"denominator":77,"numerator":2,"percentage_points":"+2.60 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_6abd58183594e22fae1475af34c1ec0277cf6a0009f12888db73e7a9f1435196","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_dd25655b484a411f612db25e334d0f23ff76fa998151734f0b04f678aefce98c","model_configuration":"o1-2024-12-17_high","observed_competition_rank":111,"organization":"OpenAI","original":{"denominator":99,"numerator":76,"percent":"76.77%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":1,"clean":{"denominator":63,"numerator":50,"percent":"79.37%"},"clean_question_n":189,"configuration_date":"2025-04-10","configuration_id":"gpqa_cfg_v1_372f86421b339917a5ea87ce041519de28881fa7f9f39ab618a6079c1668516e","delta":{"denominator":1386,"numerator":43,"percentage_points":"+3.10 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_b489e4513748213af359d6e74207b6fedf629d9f4bdfd8c5dd19fe6e5f7b9519","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_b326456b6f83bfbccf9cde7fc02a1fe96f750e01bd69cf4ed11bba768de8ba12","model_configuration":"grok-3-mini-beta_low","observed_competition_rank":111,"organization":"xAI","original":{"denominator":198,"numerator":151,"percent":"76.26%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":1,"clean":{"denominator":63,"numerator":50,"percent":"79.37%"},"clean_question_n":189,"configuration_date":"2026-08-07","configuration_id":"gpqa_cfg_v1_7db9c0fa233c93399222a3bfbac6056222f2bcacd6282d5bb91bc0f71c37ae84","delta":{"denominator":1386,"numerator":29,"percentage_points":"+2.09 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_9c0f340b80401c41bb60c544ab1d8468f1fdf56e4acbb2f32b9020be0ba25cba","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_3accbd60ebe0b166a74b437856fea71c18e59901ce28a1cc35f3c26f7135e4b4","model_configuration":"gpt-5.6-terra_none","observed_competition_rank":111,"organization":"OpenAI","original":{"denominator":22,"numerator":17,"percent":"77.27%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":8,"clean":{"denominator":189,"numerator":149,"percent":"78.84%"},"clean_question_n":189,"configuration_date":"2025-05-29","configuration_id":"gpqa_cfg_v1_678fc1ebbe85018672c472fa4980a0f10c18c89f0dccfe222b93697861ddf099","delta":{"denominator":33264,"numerator":835,"percentage_points":"+2.51 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_b193aa6a2e3321b690b8feedd221a05f415cf95482f9316833c1ed630f527370","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_d5e29f69d6f79e471f4d87e9c440a4096fd01c3d5fc73d7bf637b9b643fefbee","model_configuration":"DeepSeek-R1-0528","observed_competition_rank":114,"organization":"DeepSeek","original":{"denominator":528,"numerator":403,"percent":"76.33%"},"original_question_n":198,"protocol":"mean@8","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":1,"clean":{"denominator":189,"numerator":149,"percent":"78.84%"},"clean_question_n":189,"configuration_date":"2025-05-22","configuration_id":"gpqa_cfg_v1_9d9fcd13faf4c34aa3dfac137ba313739953d4e6cb6b34f7004fe8f8a6bc0016","delta":{"denominator":4158,"numerator":107,"percentage_points":"+2.57 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_323b108018e7aba020f9e7afa951cf32e57c3e77d59f0942c05a48d002299c99","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_bcfd0eede3b0cddcd79ced89e10a322dfd89a5df91f9b2c3cfc0e6ad6197a9b5","model_configuration":"claude-opus-4-20250514_16K","observed_competition_rank":114,"organization":"Anthropic","original":{"denominator":198,"numerator":151,"percent":"76.26%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":1,"clean":{"denominator":189,"numerator":148,"percent":"78.31%"},"clean_question_n":189,"configuration_date":"2025-05-22","configuration_id":"gpqa_cfg_v1_3b4449cd7456be9a421061071fd80145c7806c116ccfedd101c8f06b7ea16dc5","delta":{"denominator":2079,"numerator":53,"percentage_points":"+2.55 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_a574fc8089b5af19df1ac0afceca408ade5cf7e9158fe63bbd5f7983b27a3adc","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_5e7669ec1acb94109cd873958540cade2ad14d07e0554103679a173f8777a6cb","model_configuration":"claude-sonnet-4-20250514_16K","observed_competition_rank":116,"organization":"Anthropic","original":{"denominator":33,"numerator":25,"percent":"75.76%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":2,"clean":{"denominator":189,"numerator":148,"percent":"78.31%"},"clean_question_n":189,"configuration_date":"2025-12-11","configuration_id":"gpqa_cfg_v1_569cf92af26ce0effa0d7c6f3499ee8a4b59832cf43499b195482e8caf68b0e0","delta":{"denominator":2079,"numerator":53,"percentage_points":"+2.55 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_6425cfc911fd497c8b8f7eea4d2f8695e7903b9e5ff309140b4c6619b014d12b","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_79c1de5bc1713fd3813a81e8b6531d7fe800568a2b9664a42e0c8cbda907a5ba","model_configuration":"openai/gpt-oss-120b_high","observed_competition_rank":116,"organization":"OpenAI","original":{"denominator":33,"numerator":25,"percent":"75.76%"},"original_question_n":198,"protocol":"mean@2","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":1,"clean":{"denominator":189,"numerator":148,"percent":"78.31%"},"clean_question_n":189,"configuration_date":"2026-08-06","configuration_id":"gpqa_cfg_v1_5ba80b136d60d34017baf239819ca37835512b9eb474be2e9b599485c1de4c57","delta":{"denominator":2079,"numerator":53,"percentage_points":"+2.55 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_fc2703c5c2f9f81e611279817ffd0814e39c89a8abf47d038fd48d1da9ca2df1","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_49e20b164dbe1f4b3cbb7e38bf2c70d4296f9044040795b06f2964900bb91ead","model_configuration":"gemma-4-31b-it_minimal","observed_competition_rank":116,"organization":"Google DeepMind","original":{"denominator":33,"numerator":25,"percent":"75.76%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":1,"clean":{"denominator":189,"numerator":148,"percent":"78.31%"},"clean_question_n":189,"configuration_date":"2026-08-06","configuration_id":"gpqa_cfg_v1_ca66e0d6fa724848f897b1528b7c9822969050042ffe53bc36c3bad551d8bbbe","delta":{"denominator":2079,"numerator":53,"percentage_points":"+2.55 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_61b4ef5388c418bf739fe689d906ab4b0be33f09b4dba70dc8cf4912c55119d6","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_b180de987161bdbf94b237ff62e9826189aa1661c91bdfe0183489a9814f261a","model_configuration":"gemini-3.5-flash-lite_low","observed_competition_rank":116,"organization":"Google DeepMind","original":{"denominator":33,"numerator":25,"percent":"75.76%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":8,"clean":{"denominator":1512,"numerator":1177,"percent":"77.84%"},"clean_question_n":189,"configuration_date":"2025-10-30","configuration_id":"gpqa_cfg_v1_25fa96ba13ac4c3502661f5ee3958b5fead6b8ed9a5a200ed41263fcfa970d6c","delta":{"denominator":1512,"numerator":43,"percentage_points":"+2.84 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_62e3bc8326e2e952542df4ddbbbef50524d07ab02ee9b73d21c72f2262cb5854","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_d9bfe08a0212ed0f56f5bb6a95e50f5f4ed05ba6dcdfbbcdaede4306a327a80e","model_configuration":"gpt-5-mini-2025-08-07_high","observed_competition_rank":120,"organization":"OpenAI","original":{"denominator":4,"numerator":3,"percent":"75.00%"},"original_question_n":198,"protocol":"mean@8","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":1,"clean":{"denominator":9,"numerator":7,"percent":"77.78%"},"clean_question_n":189,"configuration_date":"2026-07-11","configuration_id":"gpqa_cfg_v1_06fc60637355bfe507b394a67bd3cf0abc7c538efdf32fa6908642999f53ce6d","delta":{"denominator":198,"numerator":5,"percentage_points":"+2.53 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_9a16ed69ee959ae436a9aa2b1e45b6ca842413696d7ec819e54ae9ee8b5d7042","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_931218e5efb2449dd5bab087c3c9e1a70dbb16add0778a884e14a740f9553cdc","model_configuration":"o4-mini-2025-04-16_low","observed_competition_rank":121,"organization":"OpenAI","original":{"denominator":198,"numerator":149,"percent":"75.25%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":1,"clean":{"denominator":189,"numerator":146,"percent":"77.25%"},"clean_question_n":189,"configuration_date":"2025-01-23","configuration_id":"gpqa_cfg_v1_9176a8be353931cb51348dc081ae7761a85c8a92f75a1939ae96e66f66b67305","delta":{"denominator":2079,"numerator":31,"percentage_points":"+1.49 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_d6eefa5846e1079c4320fb7305fd2ac524ddefaad2af3a84562582a79a040e9f","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_0f572fc59b5afdca713ce1314b8c70bde32f8a7f0ca91a9354cecc032867983e","model_configuration":"o1-2024-12-17_medium","observed_competition_rank":122,"organization":"OpenAI","original":{"denominator":33,"numerator":25,"percent":"75.76%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":1,"clean":{"denominator":189,"numerator":146,"percent":"77.25%"},"clean_question_n":189,"configuration_date":"2026-07-15","configuration_id":"gpqa_cfg_v1_e53c5a14f4aeac84c113c190e76b609ecf71fe6a8a08960e87cf07e538dfe203","delta":{"denominator":2079,"numerator":52,"percentage_points":"+2.50 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_d07caaae495eaf532cacb85535e3a37be795fc2d7ee2b915e44ae1b266f7e8e5","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_20469bd2762a92a967dc6393d649831dc84f0402893151ed2c8d213121629037","model_configuration":"gpt-5.4-2026-03-05_none","observed_competition_rank":122,"organization":"OpenAI","original":{"denominator":99,"numerator":74,"percent":"74.75%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":1,"clean":{"denominator":189,"numerator":145,"percent":"76.72%"},"clean_question_n":189,"configuration_date":"2026-07-20","configuration_id":"gpqa_cfg_v1_2e20ff4f87d82928ed38f791e0d054c52b489cc802e53dcdfc82bf4d51980434","delta":{"denominator":4158,"numerator":103,"percentage_points":"+2.48 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_13bf58c15c2cf11b057018e787247df5f9c2c70eb136c2c8b36b6ae8390dc068","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_f9b5365b1625ed578c5bae2c8f17551d5fc0d98b23a29cc0614830bd68b68090","model_configuration":"o1-2024-12-17_low","observed_competition_rank":124,"organization":"OpenAI","original":{"denominator":66,"numerator":49,"percent":"74.24%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":1,"clean":{"denominator":189,"numerator":145,"percent":"76.72%"},"clean_question_n":189,"configuration_date":"2026-08-06","configuration_id":"gpqa_cfg_v1_75c4a32c0454fc968aa549dba870cfd4e1bd5a1f6c823d4e434b9e24f6f77f18","delta":{"denominator":4158,"numerator":103,"percentage_points":"+2.48 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_1d7073c1b5ec278ebd0b70e087c6c9cdcc58096a06be900979e39c7b90a2a4da","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_d67b926cc255d512d885fef894902007d4e6404ec6870907d3fa016f3e806a1e","model_configuration":"gemini-3.1-flash-lite_low","observed_competition_rank":124,"organization":"Google","original":{"denominator":66,"numerator":49,"percent":"74.24%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":1,"clean":{"denominator":189,"numerator":145,"percent":"76.72%"},"clean_question_n":189,"configuration_date":"2026-08-06","configuration_id":"gpqa_cfg_v1_eb0badf81e252fdb63ecf80d40bb300712ba693f20e27eab39d39914053768f9","delta":{"denominator":4158,"numerator":103,"percentage_points":"+2.48 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_47519735add1b438c17a478ad1993733210b87670a2f431a9c3025693dd8f5c8","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_c53bfe125d212be0e5ca41fe06f7d38f12eef8451b7a84cb0ac79fc7ee4f1bf9","model_configuration":"gemini-3.5-flash-lite_minimal","observed_competition_rank":124,"organization":"Google DeepMind","original":{"denominator":66,"numerator":49,"percent":"74.24%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":16,"clean":{"denominator":1512,"numerator":1157,"percent":"76.52%"},"clean_question_n":189,"configuration_date":"2025-01-31","configuration_id":"gpqa_cfg_v1_90e5ce6b4a67252218334c529e05589fdb776c6a026b382161690d249ff27e18","delta":{"denominator":66528,"numerator":1495,"percentage_points":"+2.25 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_da0ad14552640843bcc7826c443d6dcbeff2f28482f5d200205d2ace274e80e6","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_cadf02c95b51f6a80c92e672acd6a4d0545c03ae524cb69f503648a922ce4da3","model_configuration":"o3-mini-2025-01-31_medium","observed_competition_rank":127,"organization":"OpenAI","original":{"denominator":3168,"numerator":2353,"percent":"74.27%"},"original_question_n":198,"protocol":"mean@16","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":1,"clean":{"denominator":21,"numerator":16,"percent":"76.19%"},"clean_question_n":189,"configuration_date":"2026-07-13","configuration_id":"gpqa_cfg_v1_afeeefbcbb8acb87e92f6e3238d36a6aaacc8dc9149121f1521a3cb957fcba15","delta":{"denominator":1386,"numerator":41,"percentage_points":"+2.96 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_2cca30309b582a4e3555ef0b255f847265203ef08dcbace764efec23e3893f55","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_2c05fd618e478c4fdbab86926c2327f20e857a8e3fb36b7ce506264f05559539","model_configuration":"gpt-5.2-2025-12-11_none","observed_competition_rank":128,"organization":"OpenAI","original":{"denominator":198,"numerator":145,"percent":"73.23%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":1,"clean":{"denominator":189,"numerator":143,"percent":"75.66%"},"clean_question_n":189,"configuration_date":"2025-08-05","configuration_id":"gpqa_cfg_v1_758fdeb186e7e3aa8d76746a779e63abb3f3d53282b88caf7802f648df61985a","delta":{"denominator":4158,"numerator":101,"percentage_points":"+2.43 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_ed7957e830968d6f66dac5ed25a1dde55aa7b4fc6bdfb2958e0565fcb1fca766","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_f96ec6c10dcee032be7c4c6b94888a4b94c8768f45e35c94427de90290818efd","model_configuration":"claude-opus-4-1-20250805","observed_competition_rank":129,"organization":"Anthropic","original":{"denominator":198,"numerator":145,"percent":"73.23%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":1,"clean":{"denominator":189,"numerator":143,"percent":"75.66%"},"clean_question_n":189,"configuration_date":"2026-08-06","configuration_id":"gpqa_cfg_v1_a385ae2ba85bd7d7184db8aaf5c23614c6f6b5d4da48acbb3e476fa738e7f196","delta":{"denominator":2079,"numerator":40,"percentage_points":"+1.92 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_b042b219137e076e93f11b0f62773434bba47e2808e866a349483859c342625a","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_8a50e6711db4f0b886818807f15f02136ae878f80e5d68f6e52c0a25a9c66392","model_configuration":"gemini-3.1-flash-lite_minimal","observed_competition_rank":129,"organization":"Google","original":{"denominator":99,"numerator":73,"percent":"73.74%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":1,"clean":{"denominator":189,"numerator":143,"percent":"75.66%"},"clean_question_n":189,"configuration_date":"2026-08-07","configuration_id":"gpqa_cfg_v1_bfdeb68407b430f0101f3ac3ac9ff34e4e1bcbc7034bb257d6a50df851148d7e","delta":{"denominator":4158,"numerator":101,"percentage_points":"+2.43 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_0f66c1cf64f2d84e13fca098cb05d8b0757e36a0c3961a71d6fe2ac91425bee6","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_3dbd547caea1d744783d04ac307a3e69692aefe4377b0878abe297c58c1cb12f","model_configuration":"deepseek-v4-pro_none","observed_competition_rank":129,"organization":"DeepSeek","original":{"denominator":198,"numerator":145,"percent":"73.23%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":1,"clean":{"denominator":189,"numerator":142,"percent":"75.13%"},"clean_question_n":189,"configuration_date":"2025-09-29","configuration_id":"gpqa_cfg_v1_172a5bf1e81d7ea7354117b64e82e5a863ddd3138848c51d12934571427fcfd5","delta":{"denominator":2079,"numerator":29,"percentage_points":"+1.39 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_251c6f643f2b6c39f02c07ee62c2ad525cd70fbd0586de0d985613a4b9e99080","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_666c33b98fbae48e7c49136578b6c602e384ae49094b2d2c4f0afa3756b38def","model_configuration":"claude-sonnet-4-5-20250929","observed_competition_rank":132,"organization":"Anthropic","original":{"denominator":99,"numerator":73,"percent":"73.74%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":8,"clean":{"denominator":189,"numerator":142,"percent":"75.13%"},"clean_question_n":189,"configuration_date":"2025-10-06","configuration_id":"gpqa_cfg_v1_3c5870b608f9aaf878456e51a25d8aadb6e0d8e9b1e902670d6b72045777d494","delta":{"denominator":16632,"numerator":421,"percentage_points":"+2.53 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_0233e4a8f66947e14f6135b5a7ee0d5de8ae20eee33963f722ab2c71ba784588","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_6b7ac484013d63c135c641e8c2a1e72fc0c3feae635631c5ded9e95e58e539a4","model_configuration":"qwen3-max-2025-09-23","observed_competition_rank":132,"organization":"Alibaba","original":{"denominator":792,"numerator":575,"percent":"72.60%"},"original_question_n":198,"protocol":"mean@8","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":1,"clean":{"denominator":63,"numerator":47,"percent":"74.60%"},"clean_question_n":189,"configuration_date":"2026-07-20","configuration_id":"gpqa_cfg_v1_2d16e5d29fe93a718329e4d401a08f5d17e483de216571aa6a74b0aa9f945c11","delta":{"denominator":693,"numerator":20,"percentage_points":"+2.89 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_14c867832a9a692bdf91006f3f7d4a8fba0a09e8d39e1af1a6b2d5abc320479b","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_b81787e3541ebab349fe2a001a09bce4c911e8c9b8cd5af4ab1d06e2a7912d3f","model_configuration":"gpt-5-2025-08-07_minimal","observed_competition_rank":134,"organization":"OpenAI","original":{"denominator":99,"numerator":71,"percent":"71.72%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":8,"clean":{"denominator":216,"numerator":161,"percent":"74.54%"},"clean_question_n":189,"configuration_date":"2025-08-07","configuration_id":"gpqa_cfg_v1_00b750e02386f4541fe325efb0a8449df2b72a52b8dd40a582a7a752470e3a3d","delta":{"denominator":4752,"numerator":137,"percentage_points":"+2.88 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_792c41959f3f82ccceec08b9342f234829872842cf3fcc4036e28481504fa2e3","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_d9e541024a37f5842559fb4df9b1b4f420f0361a976926db9a55a2692529c67d","model_configuration":"gpt-5-mini-2025-08-07_medium","observed_competition_rank":135,"organization":"OpenAI","original":{"denominator":1584,"numerator":1135,"percent":"71.65%"},"original_question_n":198,"protocol":"mean@8","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":1,"clean":{"denominator":27,"numerator":20,"percent":"74.07%"},"clean_question_n":189,"configuration_date":"2026-08-07","configuration_id":"gpqa_cfg_v1_53004f5f1c902a577afd76e4f9016de7963f5d869c85bb3d7906463afa273244","delta":{"denominator":54,"numerator":1,"percentage_points":"+1.85 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_7e2996f3d21e6291bdc7c3aeae14990995271215f5dccf59f770daa5482351a3","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_9d41e3a2eb0616d21676ee40f289501786e33fe43dbba87eb35ffda47da57ef6","model_configuration":"gpt-5.4-nano-2026-03-17_low","observed_competition_rank":136,"organization":"OpenAI","original":{"denominator":18,"numerator":13,"percent":"72.22%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":1,"clean":{"denominator":27,"numerator":20,"percent":"74.07%"},"clean_question_n":189,"configuration_date":"2026-08-07","configuration_id":"gpqa_cfg_v1_e8b7f10a34ef4830b8a98b69d984ed96a9afc3e9e8a7858993bef4958cb0eecc","delta":{"denominator":297,"numerator":7,"percentage_points":"+2.36 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_abb18121ff6c108f21e683268963d7b9cbb876b8a1322df5154be358e44f1485","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_0cdcc0925e07404e04451714b736e0bf13b28957076b77bb4d86f83f490e0e58","model_configuration":"gpt-5-mini-2025-08-07_minimal","observed_competition_rank":136,"organization":"OpenAI","original":{"denominator":99,"numerator":71,"percent":"71.72%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":1,"clean":{"denominator":189,"numerator":139,"percent":"73.54%"},"clean_question_n":189,"configuration_date":"2026-07-16","configuration_id":"gpqa_cfg_v1_4c3a34cfb53d1bdeb2350a0a8d43d7dd34e2878b1834dd8a615a4f6100909a4c","delta":{"denominator":4158,"numerator":97,"percentage_points":"+2.33 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_0478d6f4b4226e20b7677c937472da09020d428b98df9793e6ce5489bade9b93","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_20183cd6af48b75174c64572e9bc9e915b327f8aad814e80cbdd72db7405be62","model_configuration":"deepseek-chat","observed_competition_rank":138,"organization":"DeepSeek","original":{"denominator":66,"numerator":47,"percent":"71.21%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":1,"clean":{"denominator":189,"numerator":139,"percent":"73.54%"},"clean_question_n":189,"configuration_date":"2025-10-22","configuration_id":"gpqa_cfg_v1_b25fbe60dede1fb3425e3063a2463d869b9c4c00d4b6b0aed6903c5818c3f024","delta":{"denominator":4158,"numerator":97,"percentage_points":"+2.33 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_d5884032b0fe6618caf332eeae77e09d3029497ccb09e7d6f3d330d10b5b02ee","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_3ee2714591cbebb53903dbb1325cf6f31790b046542c5d8ba8e6e8e9465e5e00","model_configuration":"claude-haiku-4-5-20251001_32K","observed_competition_rank":138,"organization":"Anthropic","original":{"denominator":66,"numerator":47,"percent":"71.21%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":8,"clean":{"denominator":1512,"numerator":1109,"percent":"73.35%"},"clean_question_n":189,"configuration_date":"2025-06-03","configuration_id":"gpqa_cfg_v1_b6e973f9d6f6139a95d79c29885fb8d73070eadc9565d29c3978bdf06333826d","delta":{"denominator":16632,"numerator":439,"percentage_points":"+2.64 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_9e118dbe167fd2b665d992d648605798ee009bec98127fdb1c034f3eca1a59f5","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_a132b7a5bd8da5abbe730a453743f03692e12b95b6a03c9a398fe10f2073501f","model_configuration":"qwen3-235b-a22b","observed_competition_rank":140,"organization":"Alibaba","original":{"denominator":99,"numerator":70,"percent":"70.71%"},"original_question_n":198,"protocol":"mean@8","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":1,"clean":{"denominator":189,"numerator":137,"percent":"72.49%"},"clean_question_n":189,"configuration_date":"2026-08-10","configuration_id":"gpqa_cfg_v1_8d361a9663df4edd4bbe91b30f4d62bbd4be4a7a778484c5a867483645d19a6c","delta":{"denominator":4158,"numerator":53,"percentage_points":"+1.27 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_d87f5e0e4fe7e7954e6a342801578c69a887560bf781fa8f5764c150865c4845","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_15af887c85b934814a755440ae3f0d603f5fdbd1cc2a87c1c78969242d02497f","model_configuration":"glm-5.2_none","observed_competition_rank":141,"organization":"Z.ai (Zhipu AI)","original":{"denominator":66,"numerator":47,"percent":"71.21%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":8,"clean":{"denominator":1512,"numerator":1091,"percent":"72.16%"},"clean_question_n":189,"configuration_date":"2025-10-30","configuration_id":"gpqa_cfg_v1_153bcc443e2c2d97f408c94b5455fe8eded12a0a5d244c70e8fe4d7e9700b5e0","delta":{"denominator":1512,"numerator":41,"percentage_points":"+2.71 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_ca6fd8d7a3c517eaf6b033d194ee471bc1f4d5157a213767a2f4ccbe777baab6","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_936530850bcd6fd6b0da2509916a80b6b158953fefbe7488b118ccb7d763982d","model_configuration":"gpt-5-nano-2025-08-07_high","observed_competition_rank":142,"organization":"OpenAI","original":{"denominator":36,"numerator":25,"percent":"69.44%"},"original_question_n":198,"protocol":"mean@8","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":1,"clean":{"denominator":7,"numerator":5,"percent":"71.43%"},"clean_question_n":189,"configuration_date":"2025-05-22","configuration_id":"gpqa_cfg_v1_90806d19c82368364d7c0c55ac1dcd465817154c590e19fccb69699999696009","delta":{"denominator":1386,"numerator":31,"percentage_points":"+2.24 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_f1ed3cb9a6241703507b26c9843c16274c913bbf9fa5e1aca5256344444bdf58","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_87cc644071a83e463191e8a4df404665af7a8397aec7455b408e4b91a3ba015c","model_configuration":"claude-opus-4-20250514","observed_competition_rank":143,"organization":"Anthropic","original":{"denominator":198,"numerator":137,"percent":"69.19%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":1,"clean":{"denominator":189,"numerator":134,"percent":"70.90%"},"clean_question_n":189,"configuration_date":"2026-07-20","configuration_id":"gpqa_cfg_v1_68f4f26b76cc464a36f5d412acf484bac54ad4002ceef770bb5a16e84938c442","delta":{"denominator":4158,"numerator":113,"percentage_points":"+2.72 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_e77f3be95d397616315d404450f01d9a96c1a76b6ff583dca4f11237b17a8514","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_4605c22099a0980ed8cd2bfc1998083dc4dcfca3c3318b39b1f3cbf743da9771","model_configuration":"o3-mini-2025-01-31_low","observed_competition_rank":144,"organization":"OpenAI","original":{"denominator":22,"numerator":15,"percent":"68.18%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":1,"clean":{"denominator":189,"numerator":134,"percent":"70.90%"},"clean_question_n":189,"configuration_date":"2025-02-28","configuration_id":"gpqa_cfg_v1_788a0f0a26e95b58f7bad9146781e148246e45068dfba74be09ffb9e967a5f8a","delta":{"denominator":2079,"numerator":46,"percentage_points":"+2.21 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_f3fb6e2596a70a9bbb7d4d4214fdf65dad510be6e18542ee1f1911f59e8ea9e8","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_297403a07270c32fd998ca518f6cdf5668c956259038562f2be763279256da59","model_configuration":"gpt-4.5-preview-2025-02-27","observed_competition_rank":144,"organization":"OpenAI","original":{"denominator":99,"numerator":68,"percent":"68.69%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":8,"clean":{"denominator":1512,"numerator":1061,"percent":"70.17%"},"clean_question_n":189,"configuration_date":"2025-08-07","configuration_id":"gpqa_cfg_v1_cc760f480aef1f4d05a51dc48fb94f9bd8123b5a19ed8a8b623a9aa5936069ff","delta":{"denominator":16632,"numerator":457,"percentage_points":"+2.75 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_0d0d113c8d7eabc4945c34d294cc03b1a939c9dc430c2bc2e47073fa5399a810","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_0e7c910758a90f1beaa6eb551535379224127bfd2847151f0cf00a389e52e0b2","model_configuration":"gpt-5-nano-2025-08-07_medium","observed_competition_rank":146,"organization":"OpenAI","original":{"denominator":132,"numerator":89,"percent":"67.42%"},"original_question_n":198,"protocol":"mean@8","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":8,"clean":{"denominator":56,"numerator":39,"percent":"69.64%"},"clean_question_n":189,"configuration_date":"2025-04-01","configuration_id":"gpqa_cfg_v1_8377771cdc0ba4e2788fa70543be3075e24b56931201a00dc7de01ffaebc6583","delta":{"denominator":1232,"numerator":25,"percentage_points":"+2.03 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_b95c5f6b24ce76ade25a8510d7dfa4dcad0b3f60d6ff9a5419bd7149873c5c88","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_9c937b91510ffd935a34c98172ef6e8050865a7c13658a0e93aa458d42f33b6f","model_configuration":"DeepSeek-V3-0324","observed_competition_rank":147,"organization":"DeepSeek","original":{"denominator":176,"numerator":119,"percent":"67.61%"},"original_question_n":198,"protocol":"mean@8","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":8,"clean":{"denominator":189,"numerator":131,"percent":"69.31%"},"clean_question_n":189,"configuration_date":"2025-04-14","configuration_id":"gpqa_cfg_v1_934610fb54d8c86a7bea8d514dee077ee2eb5884977abbe065c70009f334e3b7","delta":{"denominator":8316,"numerator":199,"percentage_points":"+2.39 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_8742f5d043b230a117ab97f3722ca16013881b722b9406f2d2891e8c4eb57b28","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_623f1f7d3a7483eb695cbbed60452bcbf7f61b8493ba36f0ba7f480c33e2f8a3","model_configuration":"gpt-4.1-2025-04-14","observed_competition_rank":148,"organization":"OpenAI","original":{"denominator":396,"numerator":265,"percent":"66.92%"},"original_question_n":198,"protocol":"mean@8","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":1,"clean":{"denominator":189,"numerator":131,"percent":"69.31%"},"clean_question_n":189,"configuration_date":"2026-08-07","configuration_id":"gpqa_cfg_v1_ebe93e140bced2f2c9f1711124393f7b59a700e61071b4652a4827cd3e713257","delta":{"denominator":189,"numerator":5,"percentage_points":"+2.65 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_419e23c5901499f84adb13a55e1fb49886f3825bac67ccffb0c8a95e5330c540","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_eb64148b9ec84f0196e0e89910508d0d29ab7ba3969e9b7871880454d31aaca1","model_configuration":"gpt-5.1-2025-11-13_none","observed_competition_rank":148,"organization":"OpenAI","original":{"denominator":3,"numerator":2,"percent":"66.67%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":1,"clean":{"denominator":189,"numerator":130,"percent":"68.78%"},"clean_question_n":189,"configuration_date":"2025-05-22","configuration_id":"gpqa_cfg_v1_54aba346a03524e84a582d320e68f5d6e7e1ca6e1ff7a164ae903b0be1d435a6","delta":{"denominator":189,"numerator":4,"percentage_points":"+2.12 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_b4d9e290a09db97ad3f91d3eecc69fe970a85410a4d691945769b9d02d1f9683","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_f43181f3ca5b2ed334391a207f511dbf7a7f52c71f2c9895aab453ee29721dc5","model_configuration":"claude-sonnet-4-20250514","observed_competition_rank":150,"organization":"Anthropic","original":{"denominator":3,"numerator":2,"percent":"66.67%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":1,"clean":{"denominator":189,"numerator":130,"percent":"68.78%"},"clean_question_n":189,"configuration_date":"2025-06-03","configuration_id":"gpqa_cfg_v1_e965137fed2c696942a43e020de955f796ca37ab01b735427391d3454e309d79","delta":{"denominator":189,"numerator":4,"percentage_points":"+2.12 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_fab72f269f02cefca7966939c8e8f6a5bffa9549cee9e1b3090c403e466cbdc0","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_416483ba5d12e8e8f379e458032a4991a17d4b2d5e06e79b3361ff05b803e6be","model_configuration":"gemini-2.5-pro-preview-05-06","observed_competition_rank":150,"organization":"Google DeepMind","original":{"denominator":3,"numerator":2,"percent":"66.67%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":8,"clean":{"denominator":63,"numerator":43,"percent":"68.25%"},"clean_question_n":189,"configuration_date":"2025-04-08","configuration_id":"gpqa_cfg_v1_f1d46d247ebf82560a68e3bc5cf4e07c002603eaadfa51893bf7ed84564bbdbf","delta":{"denominator":3696,"numerator":47,"percentage_points":"+1.27 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_805a74e8548e777ae3cd3cfbffd8506fe94858411c008499aa834454e55cd72f","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_834c47908d3de9900df586ebe7ca68f03d829204b0ea23d8b4f496cc22bb5f13","model_configuration":"Llama-4-Maverick-17B-128E-Instruct-FP8","observed_competition_rank":152,"organization":"Meta AI","original":{"denominator":1584,"numerator":1061,"percent":"66.98%"},"original_question_n":198,"protocol":"mean@8","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":8,"clean":{"denominator":1512,"numerator":1031,"percent":"68.19%"},"clean_question_n":189,"configuration_date":"2025-04-14","configuration_id":"gpqa_cfg_v1_71cf41df6359bc5fad4222b789a003c9cc8bf989e0345e4db04c1caeaf870496","delta":{"denominator":33264,"numerator":779,"percentage_points":"+2.34 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_dd89d76ce81d26ca7c7c580858bb2a1c66866deffa835531f58e4dea0c4a0072","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_6d2891acb2ca045878b887e2da219434ac2fbbd18a3aac6c6f3d14871f555fdc","model_configuration":"gpt-4.1-mini-2025-04-14","observed_competition_rank":153,"organization":"OpenAI","original":{"denominator":1584,"numerator":1043,"percent":"65.85%"},"original_question_n":198,"protocol":"mean@8","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":8,"clean":{"denominator":1512,"numerator":1031,"percent":"68.19%"},"clean_question_n":189,"configuration_date":"2025-02-24","configuration_id":"gpqa_cfg_v1_cf79ba0cd7bfae3b0d2ce505ab46e7b0b3957b3840464c52248da9bfb033e3d1","delta":{"denominator":8316,"numerator":179,"percentage_points":"+2.15 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_eb28c28f8efad78efe685bfae2b1850b87edca5dd86b1d99cb38de1108ccfd84","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_86bdece77771e95b2bd5306d06515006efbf5c26bbb6cc432651c53d48bb77fb","model_configuration":"claude-3-7-sonnet-20250219","observed_competition_rank":153,"organization":"Anthropic","original":{"denominator":792,"numerator":523,"percent":"66.04%"},"original_question_n":198,"protocol":"mean@8","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":8,"clean":{"denominator":168,"numerator":113,"percent":"67.26%"},"clean_question_n":189,"configuration_date":"2025-04-11","configuration_id":"gpqa_cfg_v1_4e08dd7d1b1197d17bb458f074103b518735398f7272662a3a46c37f381e18c3","delta":{"denominator":5544,"numerator":103,"percentage_points":"+1.86 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_cc64f3f4ccbcb82bbbd085eff0af6626c4e907bc32ac462a1c97952a7b9f05fe","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_631b865d16ea8a60f9a73143182597786f66c30e4e4cd9ba44a27a003085179e","model_configuration":"qwq-plus","observed_competition_rank":155,"organization":"Alibaba","original":{"denominator":396,"numerator":259,"percent":"65.40%"},"original_question_n":198,"protocol":"mean@8","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":1,"clean":{"denominator":189,"numerator":127,"percent":"67.20%"},"clean_question_n":189,"configuration_date":"2025-02-07","configuration_id":"gpqa_cfg_v1_8d5c9aff7847b4d80b03824f778e301bb2f71d82a7d54bc8752a32b2d93880b9","delta":{"denominator":2079,"numerator":32,"percentage_points":"+1.54 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_b0602ac739ac95eb15086eed34884e50a032e0ce63a66984b9d8bda38e7b6c30","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_d32231861767d662305da1d39e78d304940ef6d207e9b52b3f170616bd8a2721","model_configuration":"gemini-2.0-pro-exp-02-05","observed_competition_rank":156,"organization":"Google DeepMind","original":{"denominator":99,"numerator":65,"percent":"65.66%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":1,"clean":{"denominator":189,"numerator":125,"percent":"66.14%"},"clean_question_n":189,"configuration_date":"2026-08-07","configuration_id":"gpqa_cfg_v1_6c8c1dd0bd4400f5eb27a320a475692af2d68a39af98d4f0b552fc0419ecb544","delta":{"denominator":4158,"numerator":83,"percentage_points":"+2.00 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_34dd6089244c5b77a8dfb5e171ebefb78748cafc2c60349bddeaacca71a11897","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_410175173ea89dce5f3728c07a41e8a63e7d66f1aa76fa80cf8034ca583645b2","model_configuration":"gpt-5.4-mini-2026-03-17_none","observed_competition_rank":157,"organization":"OpenAI","original":{"denominator":198,"numerator":127,"percent":"64.14%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":1,"clean":{"denominator":189,"numerator":124,"percent":"65.61%"},"clean_question_n":189,"configuration_date":"2026-08-07","configuration_id":"gpqa_cfg_v1_af5c54e09d41d150ba0a812d4c0d2dd9ded66334836f64da4a4bb102999428cd","delta":{"denominator":2079,"numerator":41,"percentage_points":"+1.97 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_791730173c974e57d98c0f611b7ce7a170f14ebdf7b1b83475cf29a12763b566","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_d79c274235ba29911b048b6b6fb91e533d94958108f9da79557a211dfcf16887","model_configuration":"gpt-5.6-luna_none","observed_competition_rank":158,"organization":"OpenAI","original":{"denominator":11,"numerator":7,"percent":"63.64%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":16,"clean":{"denominator":1512,"numerator":991,"percent":"65.54%"},"clean_question_n":189,"configuration_date":"2025-02-06","configuration_id":"gpqa_cfg_v1_e306757163f97ebb8a7e53e35a5f2c8b98039f375eba0fff85b09abd20803643","delta":{"denominator":16632,"numerator":233,"percentage_points":"+1.40 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_a8a12678ac767ad0ce614c0a52fb3cb05212df7e6fae29008ff1a1c853d066e7","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_475900f63a4c5d423cc5855258c5df8fe2c5b17f4f743308257e87deb22ca1e5","model_configuration":"gemini-2.0-flash-001","observed_competition_rank":159,"organization":"Google DeepMind,Google","original":{"denominator":198,"numerator":127,"percent":"64.14%"},"original_question_n":198,"protocol":"mean@16","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":8,"clean":{"denominator":14,"numerator":9,"percent":"64.29%"},"clean_question_n":189,"configuration_date":"2025-02-13","configuration_id":"gpqa_cfg_v1_54e29f3aa006e93a5e11a4220d4ffe0398120b3ceb40328c11661e50953226b2","delta":{"denominator":2772,"numerator":53,"percentage_points":"+1.91 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_46fcc9979861645dd49c5dfbf566c799658d885aba05a023e8ffd845d4aec810","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_522dfca6257a33cbb37330d9aa151d82132de83b96f985761e1702e486673f26","model_configuration":"o1-mini-2024-09-12_high","observed_competition_rank":160,"organization":"OpenAI","original":{"denominator":396,"numerator":247,"percent":"62.37%"},"original_question_n":198,"protocol":"mean@8","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":8,"clean":{"denominator":756,"numerator":467,"percent":"61.77%"},"clean_question_n":189,"configuration_date":"2025-10-16","configuration_id":"gpqa_cfg_v1_066a71969e005dcbca3c374019b97a0455b621a9da160724d16808d4ee7bdd9e","delta":{"denominator":16632,"numerator":215,"percentage_points":"+1.29 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_e8970f2c7575f713b293970e1f5428551b3b95a4954105e5ed533652f454d32c","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_ce73963bfa92d1e064a11add6056c83c4b69e6383db4a60c7b869d6de29398af","model_configuration":"claude-haiku-4-5-20251001","observed_competition_rank":161,"organization":"Anthropic","original":{"denominator":792,"numerator":479,"percent":"60.48%"},"original_question_n":198,"protocol":"mean@8","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":16,"clean":{"denominator":378,"numerator":233,"percent":"61.64%"},"clean_question_n":189,"configuration_date":"2025-01-17","configuration_id":"gpqa_cfg_v1_f77deb38bca0cfa2e68170980f653b63a4bea919a4a2c2475cdc2a37c7b82e48","delta":{"denominator":66528,"numerator":1423,"percentage_points":"+2.14 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_89e7ac1d13ebb3e7f1326b2e0d714604e4b41c21a83e3c8ea7485423ff0ff2e7","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_f202413019dd16e849f4fb809beb659c1a0a00d2be151135811bbfe77437672e","model_configuration":"o1-mini-2024-09-12_medium","observed_competition_rank":162,"organization":"OpenAI","original":{"denominator":3168,"numerator":1885,"percent":"59.50%"},"original_question_n":198,"protocol":"mean@16","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":8,"clean":{"denominator":756,"numerator":461,"percent":"60.98%"},"clean_question_n":189,"configuration_date":"2025-05-07","configuration_id":"gpqa_cfg_v1_202023fa5943ac9a7c62a2c1d63c955492f2ac82a4850bb837b96a4adefe8dbe","delta":{"denominator":33264,"numerator":481,"percentage_points":"+1.45 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_95c1a3eb132c968f38d952bf90086d0976307651cb68b1ff6001e60b40d893e1","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_b545ee33cdf7dee96b37c36a674516728f9bcc9057aa22a83543d7ae75160dfe","model_configuration":"mistral-medium-2505","observed_competition_rank":163,"organization":"Mistral AI","original":{"denominator":1584,"numerator":943,"percent":"59.53%"},"original_question_n":198,"protocol":"mean@8","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":1,"clean":{"denominator":27,"numerator":16,"percent":"59.26%"},"clean_question_n":189,"configuration_date":"2026-07-13","configuration_id":"gpqa_cfg_v1_69e0a76e213450c4c99586311e06c32b0188a7d080456a8a8e66f8c8bfd84fdb","delta":{"denominator":297,"numerator":5,"percentage_points":"+1.68 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_25e1ddd10af1aeefe1602f80e3c4b14035c6b3fd9fcda221ce2345f79001961e","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_7d9a6073854f79a6b73296b8bc93de50c344e9bcef48c5aeef95b308f5294e95","model_configuration":"gpt-5-nano-2025-08-07_low","observed_competition_rank":164,"organization":"OpenAI","original":{"denominator":33,"numerator":19,"percent":"57.58%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":1,"clean":{"denominator":27,"numerator":16,"percent":"59.26%"},"clean_question_n":189,"configuration_date":"2025-02-06","configuration_id":"gpqa_cfg_v1_c7c57d8f54855969165427be088ee517a83a6182a8dfb2e215a3d8ce25b8e0ef","delta":{"denominator":594,"numerator":13,"percentage_points":"+2.19 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_6577834900dfbcb589f50fc3582bde113ebf93aee5f2d1aa5d81d4e59789e406","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_cf6c49d92bc312e635cf3c7a8c96ff6a926ca11dca564059c163bc0ab00663c7","model_configuration":"gemini-2.0-flash-thinking-exp-01-21","observed_competition_rank":164,"organization":"Google DeepMind,Google","original":{"denominator":198,"numerator":113,"percent":"57.07%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":16,"clean":{"denominator":336,"numerator":197,"percent":"58.63%"},"clean_question_n":189,"configuration_date":"2025-01-17","configuration_id":"gpqa_cfg_v1_1d5d4f300a0ef5bd1a8578e997798bae60cb1a0710e2c0c8a37c01ddcaec92c5","delta":{"denominator":22176,"numerator":311,"percentage_points":"+1.40 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_fa8ef9d620dc65f782ec32bf95577bb258e8db52c4da6295af0ab70a7ee2f4d3","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_da000f42edd17dda4f211a367384b858c2cee397de1b48a51af535f720ab6df0","model_configuration":"gemini-1.5-pro-002","observed_competition_rank":166,"organization":"Google DeepMind","original":{"denominator":3168,"numerator":1813,"percent":"57.23%"},"original_question_n":198,"protocol":"mean@16","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":16,"clean":{"denominator":1008,"numerator":583,"percent":"57.84%"},"clean_question_n":189,"configuration_date":"2025-01-24","configuration_id":"gpqa_cfg_v1_2fd227a793f3f537502469f4573a2c42c5e81c9dddbb46cd8400afad108ce3ec","delta":{"denominator":22176,"numerator":289,"percentage_points":"+1.30 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_64c1c9c06c465c18a63b94f520428d5764238b73e11f5ae1981125db3b9fe397","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_c22e678b8f28d3c280fc6054b478e2a823af2cb92e1c790975684db0294cdbaa","model_configuration":"DeepSeek-V3","observed_competition_rank":167,"organization":"DeepSeek","original":{"denominator":352,"numerator":199,"percent":"56.53%"},"original_question_n":198,"protocol":"mean@16","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":8,"clean":{"denominator":756,"numerator":437,"percent":"57.80%"},"clean_question_n":189,"configuration_date":"2025-03-10","configuration_id":"gpqa_cfg_v1_d83a57d702d18c236094a9c8c5928b5c1aee75f30eb5fa16b86d87b7c6f6c312","delta":{"denominator":33264,"numerator":685,"percentage_points":"+2.06 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_1cf62a6fef04552758a0be4132c418f9d2fb9203aa472977263543983c086b56","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_bf02fd015d98b1222021573eae11fae1dd14a77e3a0dd48c197fdce39565ba3f","model_configuration":"DeepSeek-R1-Distill-Llama-70B","observed_competition_rank":168,"organization":"DeepSeek","original":{"denominator":1584,"numerator":883,"percent":"55.74%"},"original_question_n":198,"protocol":"mean@8","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":1,"clean":{"denominator":7,"numerator":4,"percent":"57.14%"},"clean_question_n":189,"configuration_date":"2026-08-06","configuration_id":"gpqa_cfg_v1_be00ebd659f5798739f50eb4ddb03eec87164876f7bc5f04ab7cadb78d6ba39c","delta":{"denominator":462,"numerator":5,"percentage_points":"+1.08 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_ee0798a79b64398cd44f82da072f043234162e67c9f3cf60c3121943b3cfda24","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_c2c9632865b7bbf8e73f7a9a723e1db9e87ac3c12347c97b537b653d6412bc9a","model_configuration":"magistral-small-2506","observed_competition_rank":169,"organization":"Mistral AI","original":{"denominator":66,"numerator":37,"percent":"56.06%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":1,"clean":{"denominator":7,"numerator":4,"percent":"57.14%"},"clean_question_n":189,"configuration_date":"2026-08-07","configuration_id":"gpqa_cfg_v1_de907b3d30ffa90340ef7c1d7ab47a458ec5f37194fde893d5354bb08a8c14fc","delta":{"denominator":63,"numerator":1,"percentage_points":"+1.59 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_4b30364c841a91549b51f0cff62105edb2a9f5186db422d8648952e345968b8d","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_620b9510b36a4677a28fe7760dd4a198397edcdce979248c60e734b8e117e9aa","model_configuration":"gpt-5.4-nano-2026-03-17_none","observed_competition_rank":169,"organization":"OpenAI","original":{"denominator":9,"numerator":5,"percent":"55.56%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":8,"clean":{"denominator":1512,"numerator":863,"percent":"57.08%"},"clean_question_n":189,"configuration_date":"2025-04-01","configuration_id":"gpqa_cfg_v1_af55e40ea3129f1d80b5438a797e0c4771653bc22e47f0d34ca14dc79968ea23","delta":{"denominator":33264,"numerator":317,"percentage_points":"+0.95 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_a74a3808884cde5a67b2377868f7515759fac615213833cf01c5285fea943c45","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_3ad649a9917c67d4ce049eecbed02a91dda453e5810ad37aa65605cac50fcbb9","model_configuration":"qwen-max-2025-01-25","observed_competition_rank":171,"organization":"Alibaba","original":{"denominator":1584,"numerator":889,"percent":"56.12%"},"original_question_n":198,"protocol":"mean@8","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":16,"clean":{"denominator":72,"numerator":41,"percent":"56.94%"},"clean_question_n":189,"configuration_date":"2025-01-31","configuration_id":"gpqa_cfg_v1_a95dfdcebc4badfda7cb23fb421a952fccefaeb6b649462fb1643f6e2ab604fc","delta":{"denominator":792,"numerator":7,"percentage_points":"+0.88 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_e2b70488b4085155fa2bcf52be361866ebc99758c5289dd751324b11e04b97af","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_5bca14b62dec3a5a55fd9c92b9e4aa5ca99ada14db4bd88a6918d8b55f1003c2","model_configuration":"phi-4","observed_competition_rank":172,"organization":"Microsoft Research","original":{"denominator":66,"numerator":37,"percent":"56.06%"},"original_question_n":198,"protocol":"mean@16","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":16,"clean":{"denominator":1008,"numerator":569,"percent":"56.45%"},"clean_question_n":189,"configuration_date":"2025-01-17","configuration_id":"gpqa_cfg_v1_d1221f2c8514da265ead360bb375cb64430d6cd83d164f77739373eae7f37879","delta":{"denominator":11088,"numerator":127,"percentage_points":"+1.15 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_b105d021c8954641ef6d8b150852735703a54dc4d76622afa303271ae0433417","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_3ec585c7244555667609b4cc8f50007cdf75bb8200ef0a2c0cd0622ce2032a60","model_configuration":"claude-3-5-sonnet-20241022","observed_competition_rank":173,"organization":"Anthropic","original":{"denominator":132,"numerator":73,"percent":"55.30%"},"original_question_n":198,"protocol":"mean@16","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":16,"clean":{"denominator":756,"numerator":421,"percent":"55.69%"},"clean_question_n":189,"configuration_date":"2025-01-17","configuration_id":"gpqa_cfg_v1_4cdcf5e7b5a7d772f20148bb6cb03e8e5e57acbe5490e9963e0df1ff0a13f315","delta":{"denominator":8316,"numerator":137,"percentage_points":"+1.65 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_634bf9f5a3db64f728934b2401ccc0d115ad69c9e938dd5eaff1ffedb1d45841","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_198d46a19024fb71e67c9f4f7c71fad94752573965683dde8f8351f8d8fda854","model_configuration":"claude-3-5-sonnet-20240620","observed_competition_rank":174,"organization":"Anthropic","original":{"denominator":198,"numerator":107,"percent":"54.04%"},"original_question_n":198,"protocol":"mean@16","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":16,"clean":{"denominator":1008,"numerator":545,"percent":"54.07%"},"clean_question_n":189,"configuration_date":"2025-01-21","configuration_id":"gpqa_cfg_v1_58b6db074aeb963d09941f6ead70727311ab03f1d3e3eb559008d9b988dccb25","delta":{"denominator":11088,"numerator":31,"percentage_points":"+0.28 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_e5b06d6b599075a92a772a091e296bfa01fd2e9190c7e4ec2e8f2ab799c5b407","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_16fb7c049832fb20f09582abe93d651ad739e000bbce6986696c5f2a37bbe2ec","model_configuration":"grok-2-1212","observed_competition_rank":175,"organization":"xAI","original":{"denominator":132,"numerator":71,"percent":"53.79%"},"original_question_n":198,"protocol":"mean@16","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":8,"clean":{"denominator":378,"numerator":199,"percent":"52.65%"},"clean_question_n":189,"configuration_date":"2025-04-08","configuration_id":"gpqa_cfg_v1_5308e01f281df009ce28185a83659d0ad19a59cdd3e55b5038c6f0b95a53d8b5","delta":{"denominator":33264,"numerator":271,"percentage_points":"+0.81 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_22da1b03bd0e0eccc48c7d5f0281927fb694a76fbf80ce2072ee1ebc84aa19ed","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_d9359bd7a357e0356b20feeb043da151e0a7ffa86c18d08eb536b4b8a960d9c7","model_configuration":"Llama-4-Scout-17B-16E-Instruct","observed_competition_rank":176,"organization":"Meta AI","original":{"denominator":1584,"numerator":821,"percent":"51.83%"},"original_question_n":198,"protocol":"mean@8","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":16,"clean":{"denominator":432,"numerator":227,"percent":"52.55%"},"clean_question_n":189,"configuration_date":"2025-01-17","configuration_id":"gpqa_cfg_v1_902ec55e0bbd019f00882dc1d375e59dbbf9bd5154a35f0daad1099813288565","delta":{"denominator":2376,"numerator":53,"percentage_points":"+2.23 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_7801ae2fa3603319120be85a5b8e63b4cdf0acd885082a516aca085cf8832ada","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_a773a87326a0a16be70727919d9511d8f0eca09dd21227358031ac9cbfafa55e","model_configuration":"o1-preview-2024-09-12","observed_competition_rank":177,"organization":"OpenAI","original":{"denominator":1584,"numerator":797,"percent":"50.32%"},"original_question_n":198,"protocol":"mean@16","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":16,"clean":{"denominator":756,"numerator":397,"percent":"52.51%"},"clean_question_n":189,"configuration_date":"2025-01-20","configuration_id":"gpqa_cfg_v1_c6f1b461149c22be3ff762bed74ced5bc575c8324790a49a899f7f672ee32cfa","delta":{"denominator":66528,"numerator":1063,"percentage_points":"+1.60 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_c2255b330f903b8fcf755fc55cefaeacf5b6dfe0959dcbe1100ec824c787166f","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_0210e91c44742be09ffb25b8c97f0a9c1eb9e1209a24a2fb3176e1c23c0ba1f7","model_configuration":"Llama-3.1-405B-Instruct","observed_competition_rank":178,"organization":"Meta AI","original":{"denominator":3168,"numerator":1613,"percent":"50.92%"},"original_question_n":198,"protocol":"mean@16","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":8,"clean":{"denominator":252,"numerator":131,"percent":"51.98%"},"clean_question_n":189,"configuration_date":"2025-02-25","configuration_id":"gpqa_cfg_v1_41455c97d103f4b57508177da6055ed522be7154caac525ae1d06daba079907f","delta":{"denominator":11088,"numerator":73,"percentage_points":"+0.66 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_93b507d8115061a7314678682e910977a0c92d617f44bceb8b13a732b5c82465","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_6870f284c264186f6933c0e05d3113748d58496f71560d088c424c042296fc32","model_configuration":"mistral-large-2411","observed_competition_rank":179,"organization":"Mistral AI","original":{"denominator":528,"numerator":271,"percent":"51.33%"},"original_question_n":198,"protocol":"mean@8","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":16,"clean":{"denominator":216,"numerator":109,"percent":"50.46%"},"clean_question_n":189,"configuration_date":"2025-01-17","configuration_id":"gpqa_cfg_v1_ee7683485330a0bb484846eff9b707715b9688adb7f1e75b90ec838b53083c08","delta":{"denominator":9504,"numerator":119,"percentage_points":"+1.25 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_67e0b61faf747c532ac6c52c154c923005bd0cbe2d60146d961de158b993737b","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_10e84a7924c260673a9bef7488f3c9b63e0f6bb27776dd6e504700e6a4a4a53a","model_configuration":"gpt-4o-2024-08-06","observed_competition_rank":180,"organization":"OpenAI","original":{"denominator":3168,"numerator":1559,"percent":"49.21%"},"original_question_n":198,"protocol":"mean@16","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":16,"clean":{"denominator":3024,"numerator":1513,"percent":"50.03%"},"clean_question_n":189,"configuration_date":"2025-01-22","configuration_id":"gpqa_cfg_v1_b36681f056f84dece01834a86b30fd4e56b779b993f6548b69eb449624c0d284","delta":{"denominator":66528,"numerator":589,"percentage_points":"+0.89 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_76be72ecb2d1012730f713f0346f63f3c49fd02d87119f948d4de67172a1ea2a","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_e8faa7b7198a65cad9ded7ad71a744e652bb936bfd5b100a03bf8c7a13b9558f","model_configuration":"qwen2.5-72b-instruct","observed_competition_rank":181,"organization":"Alibaba","original":{"denominator":352,"numerator":173,"percent":"49.15%"},"original_question_n":198,"protocol":"mean@16","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":16,"clean":{"denominator":3024,"numerator":1511,"percent":"49.97%"},"clean_question_n":189,"configuration_date":"2025-01-17","configuration_id":"gpqa_cfg_v1_5169bc0c733d68b48affc9e28c896dd620d7ff8a7a428662b2f13213317c8364","delta":{"denominator":66528,"numerator":713,"percentage_points":"+1.07 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_039b9a8b33fc8e9dcbec6b6f37a8a39e30393f5d34d30ec6a9bf90e0eec55d8c","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_243d5ed85082636e1f65f5ffc32b913ef50ce08428fcff34532ea024f39a76ae","model_configuration":"gpt-4o-2024-05-13","observed_competition_rank":182,"organization":"OpenAI","original":{"denominator":3168,"numerator":1549,"percent":"48.90%"},"original_question_n":198,"protocol":"mean@16","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":8,"clean":{"denominator":1512,"numerator":755,"percent":"49.93%"},"clean_question_n":189,"configuration_date":"2025-04-14","configuration_id":"gpqa_cfg_v1_340eb8d0df205fcfca78e8fe1bef9408f9a08baa2fbe98a058f748c1dce19c09","delta":{"denominator":33264,"numerator":335,"percentage_points":"+1.01 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_92ba7d15e4ceb81888a46ecd1f69395d19555d0b6c84d5c2d9b3f5c1f9a87953","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_687c53d044f8ad07f537ddf50a7fc464172bac7ac538c99799fdd98def9b9ab4","model_configuration":"gpt-4.1-nano-2025-04-14","observed_competition_rank":183,"organization":"OpenAI","original":{"denominator":1584,"numerator":775,"percent":"48.93%"},"original_question_n":198,"protocol":"mean@8","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":16,"clean":{"denominator":504,"numerator":251,"percent":"49.80%"},"clean_question_n":189,"configuration_date":"2025-01-22","configuration_id":"gpqa_cfg_v1_e4ead69d7e02fd150d9663a883879015c0849b66ff61f05140cf8318e45c4f6d","delta":{"denominator":22176,"numerator":173,"percentage_points":"+0.78 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_01da94af7bb743d58604ed51d08b94986a33217955dd249115f6180997b2da55","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_f01bdbc29607daac82913d046b5635bde01b1397052d7a828aee5e16548c5eb6","model_configuration":"mistral-large-2407","observed_competition_rank":184,"organization":"Mistral AI","original":{"denominator":3168,"numerator":1553,"percent":"49.02%"},"original_question_n":198,"protocol":"mean@16","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":16,"clean":{"denominator":1512,"numerator":743,"percent":"49.14%"},"clean_question_n":189,"configuration_date":"2025-02-05","configuration_id":"gpqa_cfg_v1_766ddf5c78cac9398614df89b5abd827a064cc4b8b5645bfcd7cbdf12dbc9cc8","delta":{"denominator":66528,"numerator":835,"percentage_points":"+1.26 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_0b974c884dabb85580f0492f3b4acbedb0255440c18b9267d10c8ce520d381ba","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_41e33403f86e686fd1e2a0dadcd443df22ea63ec8bc3bf8d57b05cb92dba1af2","model_configuration":"gpt-4o-2024-11-20","observed_competition_rank":185,"organization":"OpenAI","original":{"denominator":3168,"numerator":1517,"percent":"47.89%"},"original_question_n":198,"protocol":"mean@16","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":1,"clean":{"denominator":189,"numerator":92,"percent":"48.68%"},"clean_question_n":189,"configuration_date":"2026-07-13","configuration_id":"gpqa_cfg_v1_e303a40eed6943f0cc681f7368cff1f8202b2615ff9e972a0f6b2ec5c323afbc","delta":{"denominator":2079,"numerator":4,"percentage_points":"+0.19 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_52b9ebb484e2f8e3269183cfeda70aa34c7ff9f273833761779c21b713a3d7a5","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_43384103fa2c391d8438da506a04ac275685979836549369ee26c4e24809aa64","model_configuration":"gpt-5-nano-2025-08-07_minimal","observed_competition_rank":186,"organization":"OpenAI","original":{"denominator":33,"numerator":16,"percent":"48.48%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":8,"clean":{"denominator":1512,"numerator":731,"percent":"48.35%"},"clean_question_n":189,"configuration_date":"2025-04-07","configuration_id":"gpqa_cfg_v1_5b80556138a24f63b8f2cb80dccbea6656b707f83aedcd63a5cfe0c1d618c3d1","delta":{"denominator":2079,"numerator":5,"percentage_points":"+0.24 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_5eb40b4a1fc21a66c876f059bb93f6f5a07dc7420d4b52b8bbf2bab9f19f06b2","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_0e53e19644c87dc0e84c9e84ee9e3527e1f1f97f4dcd02d8d939efb7eb0401e0","model_configuration":"qwen-plus-2025-01-25","observed_competition_rank":187,"organization":"Alibaba","original":{"denominator":264,"numerator":127,"percent":"48.11%"},"original_question_n":198,"protocol":"mean@8","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":16,"clean":{"denominator":27,"numerator":13,"percent":"48.15%"},"clean_question_n":189,"configuration_date":"2025-01-22","configuration_id":"gpqa_cfg_v1_346701427f0d44ced8ca606e9855b570bb07e6ef55855f0a6332eed79ea25189","delta":{"denominator":9504,"numerator":67,"percentage_points":"+0.70 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_eacdbe9bc45ea9592ffc838201b2238e93db76ab19475c10e241d5443ae49f0a","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_50eebed0c14408666088d581f43d6c5da73019ea2ec8b715b708b1f9e9f97afc","model_configuration":"Llama-3.3-70B-Instruct","observed_competition_rank":188,"organization":"Meta AI","original":{"denominator":352,"numerator":167,"percent":"47.44%"},"original_question_n":198,"protocol":"mean@16","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":1,"clean":{"denominator":27,"numerator":13,"percent":"48.15%"},"clean_question_n":189,"configuration_date":"2026-08-06","configuration_id":"gpqa_cfg_v1_a0e72767f9478bbe4b04b9a3bea317ca374e27b8e4bbf00393cac2b87d1a2775","delta":{"denominator":594,"numerator":13,"percentage_points":"+2.19 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_a712406e21d87161483d38e2436f8659ea8517a92ddd1b5e0d62cd6c7041350f","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_00f055aabb5434ab877a0514d4d53738943a2425d8671c732d841e26ebe63241","model_configuration":"gpt-oss-20b_high","observed_competition_rank":188,"organization":"OpenAI","original":{"denominator":198,"numerator":91,"percent":"45.96%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":16,"clean":{"denominator":3024,"numerator":1453,"percent":"48.05%"},"clean_question_n":189,"configuration_date":"2025-01-17","configuration_id":"gpqa_cfg_v1_a655fc2ed8e0041b02a2ea22c580e50c6da8f701a9667563da361120fddcdab2","delta":{"denominator":66528,"numerator":487,"percentage_points":"+0.73 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_cc2237da439b936769ccc5f28f02a2e51f625ff41cabd8e9752ff05e864a806c","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_046386ed49e6001d37de1c91c358d44e7aa2d07114feaf85a7344d39d3391e62","model_configuration":"gemini-1.5-flash-002","observed_competition_rank":190,"organization":"Google DeepMind","original":{"denominator":3168,"numerator":1499,"percent":"47.32%"},"original_question_n":198,"protocol":"mean@16","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":8,"clean":{"denominator":252,"numerator":121,"percent":"48.02%"},"clean_question_n":189,"configuration_date":"2025-03-18","configuration_id":"gpqa_cfg_v1_fe008500b306f51fcb6fd66c7d472ecf21625ed009686be7a81709ebd6aad0c7","delta":{"denominator":924,"numerator":5,"percentage_points":"+0.54 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_d88b3fde412bc3daef1d1dbaac2e7984bf68c0f93a18e75fcab47d9a1f1dc75f","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_d3f22782b417834252665ddb1102c1fc18877445b8152cbacf1ca1140fe4664d","model_configuration":"mistral-small-2503","observed_competition_rank":191,"organization":"Mistral AI","original":{"denominator":99,"numerator":47,"percent":"47.47%"},"original_question_n":198,"protocol":"mean@8","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":16,"clean":{"denominator":378,"numerator":181,"percent":"47.88%"},"clean_question_n":189,"configuration_date":"2025-01-17","configuration_id":"gpqa_cfg_v1_59886bda1c8504af950b9962ec90464beff5b80e797a4d747fee5b15546d0674","delta":{"denominator":33264,"numerator":241,"percentage_points":"+0.72 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_2877769eccb658339c928b8169c2b496940c7a5cf8b027fa9d9c9dad4f96b44e","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_cf36144f7e48a23a6a2309dfceda87c46953316341453d3108263f37f31d00e3","model_configuration":"claude-3-opus-20240229","observed_competition_rank":192,"organization":"Anthropic","original":{"denominator":176,"numerator":83,"percent":"47.16%"},"original_question_n":198,"protocol":"mean@16","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":8,"clean":{"denominator":756,"numerator":355,"percent":"46.96%"},"clean_question_n":189,"configuration_date":"2025-02-27","configuration_id":"gpqa_cfg_v1_e0a3d196b9052321e7008c698e73f936a46f53cfdf612305492866420090cc21","delta":{"denominator":16632,"numerator":61,"percentage_points":"+0.37 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_7190e4ed2bc64cec4c9dad819a2bf26b434da72593287bebdbaff37a24feeed2","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_614f4386bf81f4cf532eb2cc0564908bc03e9f16ec017282de07dbe80fc46c4f","model_configuration":"gpt-4-turbo-2024-04-09","observed_competition_rank":193,"organization":"OpenAI","original":{"denominator":88,"numerator":41,"percent":"46.59%"},"original_question_n":198,"protocol":"mean@8","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":16,"clean":{"denominator":756,"numerator":355,"percent":"46.96%"},"clean_question_n":189,"configuration_date":"2025-01-22","configuration_id":"gpqa_cfg_v1_e8a07532d90482fd2050087865a6e3e30cf6a9ffb8a09595ff45e6be66586c51","delta":{"denominator":33264,"numerator":227,"percentage_points":"+0.68 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_7328c8f7177f793e15d9d53f2624e436d745bbff10e6af514a4b2640a127bd81","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_60cc48b4832f2b78d2284b10200cf111ff6b2a011c4185c570b5f9f8e2ac61eb","model_configuration":"Llama-3.1-Tulu-3-70B-DPO","observed_competition_rank":193,"organization":"Allen Institute for AI,University of Washington","original":{"denominator":1584,"numerator":733,"percent":"46.28%"},"original_question_n":198,"protocol":"mean@16","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":16,"clean":{"denominator":504,"numerator":235,"percent":"46.63%"},"clean_question_n":189,"configuration_date":"2025-01-30","configuration_id":"gpqa_cfg_v1_0aecd21ee0bc485147dfb311f84fdcd13ab5be37a49413da6200f0e3a3f8a9e4","delta":{"denominator":924,"numerator":5,"percentage_points":"+0.54 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_46b41b2c72fc6db332d75c94d8c8133134f7350e46ceb3cfd161d0cb03b21859","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_bb1e2d5bf137420f1d4aca14a95963c00a74060207679f087ecccfed1bb2f326","model_configuration":"qwen2.5-32b-instruct","observed_competition_rank":195,"organization":"Alibaba","original":{"denominator":792,"numerator":365,"percent":"46.09%"},"original_question_n":198,"protocol":"mean@16","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":8,"clean":{"denominator":1512,"numerator":703,"percent":"46.49%"},"clean_question_n":189,"configuration_date":"2025-03-10","configuration_id":"gpqa_cfg_v1_fb3188e98dcd04325c815261985bb5384ea7240ee9d0ef804de82473aede6bc9","delta":{"denominator":16632,"numerator":299,"percentage_points":"+1.80 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_f769a1bb47fe0a249f6a61005b7ad9fa5cb56532a7c6c2cde29b07fc99a78462","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_deb5a2dc0e3d18d13b3f704e153198f623e64bbe63a62ea3e8bba380e8e6e30b","model_configuration":"DeepSeek-R1-Distill-Qwen-14B","observed_competition_rank":196,"organization":"DeepSeek","original":{"denominator":132,"numerator":59,"percent":"44.70%"},"original_question_n":198,"protocol":"mean@8","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":16,"clean":{"denominator":28,"numerator":13,"percent":"46.43%"},"clean_question_n":189,"configuration_date":"2025-01-17","configuration_id":"gpqa_cfg_v1_e9fe429ba7834a25f5cf0d8f2deb496f1db6fe7b7552467a7b880fdd61f18cc6","delta":{"denominator":22176,"numerator":125,"percentage_points":"+0.56 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_75308a9a4ce0f186eaeb19b265a31da6d64b5697df4ec64bfcbe3d350c15a145","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_3f3dfa8d602ab38ffad53d167bf75d6e6d2bc864ef55a1621a34e1c62ea15b30","model_configuration":"gemini-1.5-pro-001","observed_competition_rank":197,"organization":"Google DeepMind","original":{"denominator":3168,"numerator":1453,"percent":"45.86%"},"original_question_n":198,"protocol":"mean@16","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":16,"clean":{"denominator":432,"numerator":199,"percent":"46.06%"},"clean_question_n":189,"configuration_date":"2025-01-30","configuration_id":"gpqa_cfg_v1_3d7b3a0e3e36b74bdde08e2c386e942a53b522bd7c61c1151e6dc007e6e9f001","delta":{"denominator":9504,"numerator":73,"percentage_points":"+0.77 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_272608934e92117c602d64514128c1867a07c1b57115700655d642e1fb197f2b","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_56ada95164ff407147cf7b22059b1b336410612ab62b0e4d18ba32d836c47c5d","model_configuration":"mistral-small-2501","observed_competition_rank":198,"organization":"Mistral AI","original":{"denominator":3168,"numerator":1435,"percent":"45.30%"},"original_question_n":198,"protocol":"mean@16","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":1,"clean":{"denominator":189,"numerator":85,"percent":"44.97%"},"clean_question_n":189,"configuration_date":"2026-08-06","configuration_id":"gpqa_cfg_v1_dfcacd48f5945f1701b6202adec1a676dc8eb0e19b95f390dfe26319840ae1d5","delta":{"denominator":4158,"numerator":43,"percentage_points":"+1.03 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_147dc8f413f53a986ea0be696787aa4a7bcc8f7c81f149e1b7e184866ca9c683","evidence_class":"direct_strict","evidence_variant_id":"gpqa_evd_v1_fa4d7e040bc5714277737efe82f63c1f88d40ca91d4d26a189907fdb1fd9afee","model_configuration":"gemma-3-27b-it","observed_competition_rank":199,"organization":"Google DeepMind","original":{"denominator":66,"numerator":29,"percent":"43.94%"},"original_question_n":198,"protocol":"mean@1","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"strict_complete"},{"attempts_per_question":16,"clean":{"denominator":3024,"numerator":1357,"percent":"44.87%"},"clean_question_n":189,"configuration_date":"2025-01-20","configuration_id":"gpqa_cfg_v1_536a3d475a86a0bde0d95ab2fea86e8ab597b125c21175652a1b26a307d66765","delta":{"denominator":33264,"numerator":227,"percentage_points":"+0.68 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_b7988379e8ab98297f032f9b27c27ce6660a7843e128165e84d42f0a6c76d710","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_b16503dd683d625d37a83787926aecad20fd1027038dd1d52b14d7274d1e0d65","model_configuration":"Llama-3.1-70B-Instruct","observed_competition_rank":200,"organization":"Meta AI","original":{"denominator":396,"numerator":175,"percent":"44.19%"},"original_question_n":198,"protocol":"mean@16","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":16,"clean":{"denominator":3024,"numerator":1339,"percent":"44.28%"},"clean_question_n":189,"configuration_date":"2025-01-21","configuration_id":"gpqa_cfg_v1_b8507ecd7db462de6e723c0dd6cefa055697adc5afe4eaa772a1d3f43c291371","delta":{"denominator":33264,"numerator":281,"percentage_points":"+0.84 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_1c4918f08653f9b887f57bafc02a28a088ec4633d22db99727ad4d3740d00e21","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_5a1006a45145e6d1f13ee773ce37108bd2b070ba59b8c252e336a15f9987b3a5","model_configuration":"WizardLM-2-8x22B","observed_competition_rank":201,"organization":"Microsoft","original":{"denominator":99,"numerator":43,"percent":"43.43%"},"original_question_n":198,"protocol":"mean@16","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":16,"clean":{"denominator":3024,"numerator":1313,"percent":"43.42%"},"clean_question_n":189,"configuration_date":"2025-01-17","configuration_id":"gpqa_cfg_v1_bd2829d168e72420d4610fdbbbcdd955033646500d77a719befb520b54bcc796","delta":{"denominator":189,"numerator":2,"percentage_points":"+1.06 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_556475bf264b636c3d54229a7e82a9bba4f3848974f830dff8ccdb7a312b4647","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_946bc4c5c62e8e724c128e71783e863c64910c7f628307b210ad614581401316","model_configuration":"gpt-4-1106-preview","observed_competition_rank":202,"organization":"OpenAI","original":{"denominator":144,"numerator":61,"percent":"42.36%"},"original_question_n":198,"protocol":"mean@16","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":16,"clean":{"denominator":189,"numerator":82,"percent":"43.39%"},"clean_question_n":189,"configuration_date":"2025-01-17","configuration_id":"gpqa_cfg_v1_8c7fa3c8345975553ad175f719559715fdda562d71eb9822dec2cf33e6dfeb1e","delta":{"denominator":66528,"numerator":745,"percentage_points":"+1.12 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_fd1b02832b3a497ea2ddebabaf67d89d1969262cab76cc266f44c5f7261faefa","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_70611083db50c82b95c428288e25680fc6e62ad8850949f6b3c45bb650df36d3","model_configuration":"gpt-4-0125-preview","observed_competition_rank":203,"organization":"OpenAI","original":{"denominator":3168,"numerator":1339,"percent":"42.27%"},"original_question_n":198,"protocol":"mean@16","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":8,"clean":{"denominator":189,"numerator":79,"percent":"41.80%"},"clean_question_n":189,"configuration_date":"2025-04-07","configuration_id":"gpqa_cfg_v1_a8f7b12d9b5ebb9fdec3e12f1320eef478aff93bb92584ed2185df0ebaf28be5","delta":{"denominator":16632,"numerator":1,"percentage_points":"+0.01 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_623be4cb923c894d0ba39c8b55b80262820038b25fbf64edc1e79d9b68d4eb4e","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_af128ee2d2ac65a058ec23e7de20431228f2a59971664236ea6299d00df1d105","model_configuration":"qwen-turbo-2024-11-01","observed_competition_rank":204,"organization":"Alibaba","original":{"denominator":792,"numerator":331,"percent":"41.79%"},"original_question_n":198,"protocol":"mean@8","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":16,"clean":{"denominator":3024,"numerator":1259,"percent":"41.63%"},"clean_question_n":189,"configuration_date":"2025-01-22","configuration_id":"gpqa_cfg_v1_b469234b3adc87ce38b386bcaf40c26f1aa236300b162eae3f193b325d11ed95","delta":{"denominator":33264,"numerator":199,"percentage_points":"+0.60 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_18ede0f30376d27589270d01b893d1a9e6ab9daa2988644267e40983cfaacd54","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_8e628442c92f56910bae1767681a204217285335eb2d8ed4366f7857d08a5496","model_configuration":"Llama-3.2-90B-Vision-Instruct","observed_competition_rank":205,"organization":"Meta AI","original":{"denominator":792,"numerator":325,"percent":"41.04%"},"original_question_n":198,"protocol":"mean@16","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":16,"clean":{"denominator":1008,"numerator":419,"percent":"41.57%"},"clean_question_n":189,"configuration_date":"2025-01-21","configuration_id":"gpqa_cfg_v1_58e2ae82542f35fedd02586ab2a506dc9dd9a92ad9f1913f2bcb0393aeaf603b","delta":{"denominator":22176,"numerator":223,"percentage_points":"+1.01 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_adae307af1764f3d1c10f241782323fb273b930e20aebda1a4c6c1c192aa7cd3","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_248a16f13c340919d1ee009aa68164ebe4278b515dfce8051ba5c3d9d1c6585d","model_configuration":"Meta-Llama-3-70B-Instruct","observed_competition_rank":206,"organization":"Meta AI","original":{"denominator":3168,"numerator":1285,"percent":"40.56%"},"original_question_n":198,"protocol":"mean@16","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":16,"clean":{"denominator":3024,"numerator":1249,"percent":"41.30%"},"clean_question_n":189,"configuration_date":"2025-01-17","configuration_id":"gpqa_cfg_v1_56d66d76f7fd4532bd5045af003c34d1f07f98b76657161a697fa41101524e25","delta":{"denominator":8316,"numerator":59,"percentage_points":"+0.71 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_441d4e54e4f5b3ef3a0baca49e9b9ed90456b39bb8196a7eb27954c07c7559fd","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_c0532c6558b99168a64b9955a9315f8b6e21f0ba2d7b2f3af63334af7b628c99","model_configuration":"claude-3-sonnet-20240229","observed_competition_rank":207,"organization":"Anthropic","original":{"denominator":1584,"numerator":643,"percent":"40.59%"},"original_question_n":198,"protocol":"mean@16","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":16,"clean":{"denominator":3024,"numerator":1237,"percent":"40.91%"},"clean_question_n":189,"configuration_date":"2025-01-20","configuration_id":"gpqa_cfg_v1_6970bd17ac70725a425ec6d86eb63a2d2c878ee8acefa59ba35a8efaa4180714","delta":{"denominator":33264,"numerator":41,"percentage_points":"+0.12 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_cbaf44835cd40df0459141c77d8487c65d1c16f08df2868ebd3051fef26f126b","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_c257c84e036a375f8ed1ef5dadbd25a685b4c726883640acb3d45e7485443089","model_configuration":"qwen2-72b-instruct","observed_competition_rank":208,"organization":"Alibaba","original":{"denominator":792,"numerator":323,"percent":"40.78%"},"original_question_n":198,"protocol":"mean@16","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":16,"clean":{"denominator":1008,"numerator":407,"percent":"40.38%"},"clean_question_n":189,"configuration_date":"2025-01-27","configuration_id":"gpqa_cfg_v1_1d957091393fa4b70f270be10a21999490fcaeae1de9aa2a1d5f9a5196c80ea0","delta":{"denominator":22176,"numerator":1,"percentage_points":"0.00 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_5d709ffdeedc728b74e4f1e90b1d5e6f71ee0d50aff14b612c3554a41b38afbe","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_f35e032ebccb00ab309725ac6ba4027a25988d59ec638705940d71460ce3c2e2","model_configuration":"gemini-1.5-flash-001","observed_competition_rank":209,"organization":"Google DeepMind","original":{"denominator":3168,"numerator":1279,"percent":"40.37%"},"original_question_n":198,"protocol":"mean@16","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":16,"clean":{"denominator":378,"numerator":149,"percent":"39.42%"},"clean_question_n":189,"configuration_date":"2025-01-22","configuration_id":"gpqa_cfg_v1_4c7af7da68de4f30e4d7cddcc9db1fbbf77833670cae85770afbb589903f3414","delta":{"denominator":16632,"numerator":109,"percentage_points":"+0.66 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_e2029b0e278266d4c4aad2243a7526b13fae5e7358fcf87c16fa509cc40daf35","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_f50d63ba07fedab826a9217dd7cdea0d41bc467f08e51da4d9ab10df4d5ece40","model_configuration":"mistral-large-2402","observed_competition_rank":210,"organization":"Mistral AI","original":{"denominator":792,"numerator":307,"percent":"38.76%"},"original_question_n":198,"protocol":"mean@16","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":8,"clean":{"denominator":1512,"numerator":589,"percent":"38.96%"},"clean_question_n":189,"configuration_date":"2025-03-12","configuration_id":"gpqa_cfg_v1_28ad4b722e7c76693ee321fb599eaa8bcce2126afeffed3cb5a53f356a229340","delta":{"denominator":16632,"numerator":137,"percentage_points":"+0.82 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_fb34c3033b6936e64041a9176019cecb087b661a1a9213f71544afacf98da583","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_d7db75f0655adbf2456287bb2f2328e08012b49ca8e9325ed31716ad712ae1fc","model_configuration":"claude-3-5-haiku-20241022","observed_competition_rank":211,"organization":"Anthropic","original":{"denominator":396,"numerator":151,"percent":"38.13%"},"original_question_n":198,"protocol":"mean@8","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":16,"clean":{"denominator":1512,"numerator":587,"percent":"38.82%"},"clean_question_n":189,"configuration_date":"2025-01-17","configuration_id":"gpqa_cfg_v1_d212f3b7012127430e247ff410264d8f3ad2193551b8da2fd4265bff806d73fa","delta":{"denominator":66528,"numerator":733,"percentage_points":"+1.10 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_8fc819f56f02502eeca9699434f3dd5b84bdb0616d07c0a012272c7a721786b1","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_2e9f5bc60a43efb1e18bff823dd3eb43757ce1e13a698d8a5a9d24d3a785c77f","model_configuration":"gpt-4o-mini-2024-07-18","observed_competition_rank":212,"organization":"OpenAI","original":{"denominator":3168,"numerator":1195,"percent":"37.72%"},"original_question_n":198,"protocol":"mean@16","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":16,"clean":{"denominator":3024,"numerator":1157,"percent":"38.26%"},"clean_question_n":189,"configuration_date":"2025-01-21","configuration_id":"gpqa_cfg_v1_322dfa16da9ed38170bad2a7f7bdc94967cda750fa07cddee0c558f1bd8d506e","delta":{"denominator":66528,"numerator":527,"percentage_points":"+0.79 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_39c4ce7df9b3ade8c5062cd6b5b7991617fd956fe43c7080d17910d6b97a05eb","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_224469e684cd3105f7693029831040d119a347ce316175ad3a01c2e87bd2dd3f","model_configuration":"Hermes-2-Theta-Llama-3-70B","observed_competition_rank":213,"organization":"Nous Research,Arcee AI","original":{"denominator":3168,"numerator":1187,"percent":"37.47%"},"original_question_n":198,"protocol":"mean@16","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":16,"clean":{"denominator":756,"numerator":275,"percent":"36.38%"},"clean_question_n":189,"configuration_date":"2025-01-17","configuration_id":"gpqa_cfg_v1_8dc9a927af2441ac3a54b771539215cedc19d0bdaf6864c0faff20a39f9e062f","delta":{"denominator":33264,"numerator":25,"percentage_points":"+0.08 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_95795a18eae1390656174f941596019018183dc96fa60299a2ea18a566d4d000","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_3f0bb4738ba6e2ff2de8408943325693e40959a162efdddf5977f4860aa470c3","model_configuration":"claude-3-haiku-20240307","observed_competition_rank":214,"organization":"Anthropic","original":{"denominator":1584,"numerator":575,"percent":"36.30%"},"original_question_n":198,"protocol":"mean@16","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":16,"clean":{"denominator":378,"numerator":137,"percent":"36.24%"},"clean_question_n":189,"configuration_date":"2025-01-17","configuration_id":"gpqa_cfg_v1_3706e715c29d8cc087e5eea8fcdcfc37f9f912420fe1550b3d630f10b709d0cc","delta":{"denominator":16632,"numerator":-41,"percentage_points":"-0.25 pp"},"delta_direction":"decrease","evaluation_run_id":"gpqa_run_v1_fc5488cb3b2dd15c9ca55eaa3ca5e1d9a9dd45a04f478f2b5e3ee33cfdf7d6e4","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_8fbb01bae2958144e807395932fd73caf9656f6bc2e72fb5df5233603ad21814","model_configuration":"gemma-2-27b-it","observed_competition_rank":215,"organization":"Google DeepMind","original":{"denominator":792,"numerator":289,"percent":"36.49%"},"original_question_n":198,"protocol":"mean@16","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":8,"clean":{"denominator":1512,"numerator":545,"percent":"36.04%"},"clean_question_n":189,"configuration_date":"2025-10-23","configuration_id":"gpqa_cfg_v1_58d9552106e8f5dfbcb2f79f29d67a6ee305a8dc9f8492f63ddf624493d7530a","delta":{"denominator":4158,"numerator":13,"percentage_points":"+0.31 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_926aed6b613d9dd91e10f3a53004f8390136ded8ff7ae56eaa721656e8d8b8aa","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_d4b7ed220a632326d3c67ac1fa54c4e79d8dc6a73dabaa9877b432abf48f3eb6","model_configuration":"gpt-4-0314","observed_competition_rank":216,"organization":"OpenAI","original":{"denominator":792,"numerator":283,"percent":"35.73%"},"original_question_n":198,"protocol":"mean@8","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":16,"clean":{"denominator":1512,"numerator":523,"percent":"34.59%"},"clean_question_n":189,"configuration_date":"2025-01-21","configuration_id":"gpqa_cfg_v1_c030a6e64a8b74f6e73831a3aae768dc618173ab7994e9387b5f3a60cd882de0","delta":{"denominator":2079,"numerator":13,"percentage_points":"+0.63 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_4d4f8e3135650879f10070cd2680e14f6a49e663a66f52da7d43541d4a221050","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_71700c48dc2e1abbfec61082f8ae34ba9d3037412e9b9ae8348008be452dbf27","model_configuration":"gemini-1.0-pro-001","observed_competition_rank":217,"organization":"Google DeepMind","original":{"denominator":792,"numerator":269,"percent":"33.96%"},"original_question_n":198,"protocol":"mean@16","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":16,"clean":{"denominator":108,"numerator":37,"percent":"34.26%"},"clean_question_n":189,"configuration_date":"2025-01-22","configuration_id":"gpqa_cfg_v1_29e5aee9510b318690e575f5c2a1337b657fda114d776a3b4bbd40440424339b","delta":{"denominator":9504,"numerator":19,"percentage_points":"+0.20 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_e5477e5a40319b819a2129e0271fad39918c8b13f8ef58231014c64fcde2340e","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_436ff58a85ef500c61572711dda6fbcb0b8acae285c087098a98f0597eec8949","model_configuration":"open-mixtral-8x22b","observed_competition_rank":218,"organization":"Mistral AI","original":{"denominator":3168,"numerator":1079,"percent":"34.06%"},"original_question_n":198,"protocol":"mean@16","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":16,"clean":{"denominator":1512,"numerator":517,"percent":"34.19%"},"clean_question_n":189,"configuration_date":"2025-01-17","configuration_id":"gpqa_cfg_v1_bb7b9a95d5cff8aa62bb04540cc6a3c48427cc5b28e23bde5a188484ba9cda5d","delta":{"denominator":33264,"numerator":-155,"percentage_points":"-0.47 pp"},"delta_direction":"decrease","evaluation_run_id":"gpqa_run_v1_ff9ed21ccc61d209594a1e016635d3088037d63ddac362a0e43bef48729d5892","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_759f3f6f4b5a2cf6f8c0f0854f626fd30c94a331e3abf1b238813400e9af6b72","model_configuration":"claude-2.0","observed_competition_rank":219,"organization":"Anthropic","original":{"denominator":176,"numerator":61,"percent":"34.66%"},"original_question_n":198,"protocol":"mean@16","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":16,"clean":{"denominator":3024,"numerator":1033,"percent":"34.16%"},"clean_question_n":189,"configuration_date":"2025-01-23","configuration_id":"gpqa_cfg_v1_724ee2f69b618a701e0d80941d264c74cb5c2349b4b958a51c605beca4d829c3","delta":{"denominator":16632,"numerator":43,"percentage_points":"+0.26 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_626b81dd77f63dd21786d87d428c9e61aed4851ee7dae2c85490a0fbb875d735","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_4720948ed9461d98b96792c144585b799f71e0395f59f66fa2bc14271dda35af","model_configuration":"Eurus-2-7B-PRIME","observed_competition_rank":220,"organization":"Tsinghua University,University of Illinois Urbana-Champaign (UIUC),Shanghai AI Lab,Peking University,Shanghai Jiao Tong University,CUHK Shenzhen Research Institute","original":{"denominator":528,"numerator":179,"percent":"33.90%"},"original_question_n":198,"protocol":"mean@16","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":16,"clean":{"denominator":756,"numerator":253,"percent":"33.47%"},"clean_question_n":189,"configuration_date":"2025-01-17","configuration_id":"gpqa_cfg_v1_f2ec9c6f943036ef8ca6bef0edc2ea301077976070fe65d840676bbb44703a53","delta":{"denominator":16632,"numerator":85,"percentage_points":"+0.51 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_6708f641f7b1c2453853d0c0f5b44fce92f7e1fe8bbaf674c8ac8b68e3ebd355","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_18558b809255892beb305d1594a920bffd5ebadad5299e140cf625a0a4b32be5","model_configuration":"claude-2.1","observed_competition_rank":221,"organization":"Anthropic","original":{"denominator":88,"numerator":29,"percent":"32.95%"},"original_question_n":198,"protocol":"mean@16","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":16,"clean":{"denominator":1512,"numerator":505,"percent":"33.40%"},"clean_question_n":189,"configuration_date":"2025-01-21","configuration_id":"gpqa_cfg_v1_68be4e024b354a452f5a19040775c2d5aecb864abe7845e7832d52a990885029","delta":{"denominator":33264,"numerator":169,"percentage_points":"+0.51 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_feac116dd3feb7570ec416738dea26a2b6c9d330ac69ac2286da20bc363fda19","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_f6ff488a8cf0932cbd19204899fc2efb3e377f74078010108166cab31c2f80cf","model_configuration":"dbrx-instruct","observed_competition_rank":222,"organization":"Databricks","original":{"denominator":1584,"numerator":521,"percent":"32.89%"},"original_question_n":198,"protocol":"mean@16","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":8,"clean":{"denominator":1512,"numerator":499,"percent":"33.00%"},"clean_question_n":189,"configuration_date":"2025-03-12","configuration_id":"gpqa_cfg_v1_f37332153c1ff0005df8b2ece84a9097d786a435e9046ba52befebedb2adfd3e","delta":{"denominator":2079,"numerator":1,"percentage_points":"+0.05 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_9e34e9e4b250171d57b71fa15eb1d53e2777c0e37c57f5e790702d34506d435b","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_c47972f8303d2007196ba7f5312bc24202247bddab4ecdffe534d98886e18363","model_configuration":"gemini-1.5-flash-8b-001","observed_competition_rank":223,"organization":"Google DeepMind","original":{"denominator":88,"numerator":29,"percent":"32.95%"},"original_question_n":198,"protocol":"mean@8","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":16,"clean":{"denominator":1008,"numerator":325,"percent":"32.24%"},"clean_question_n":189,"configuration_date":"2025-01-21","configuration_id":"gpqa_cfg_v1_cf9f9998388c8eddb101a1c6d6fdf01748173bb032842e230ac263e40899daa5","delta":{"denominator":22176,"numerator":59,"percentage_points":"+0.27 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_cc195c4dda12a63110db2e3a9d13dc8d93244f1708be103f58ff08786f43648f","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_5ebc7f414be92b8496480dd45cf9a6128627b0f00a0a0ec5025a55245d488e65","model_configuration":"Yi-1.5-34B-Chat","observed_competition_rank":224,"organization":"01.AI","original":{"denominator":3168,"numerator":1013,"percent":"31.98%"},"original_question_n":198,"protocol":"mean@16","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":16,"clean":{"denominator":3024,"numerator":947,"percent":"31.32%"},"clean_question_n":189,"configuration_date":"2025-01-17","configuration_id":"gpqa_cfg_v1_7671edcdc19c9d32865c60525cb98415cb4da59e1b67ca91aa8f1c27e9a823bf","delta":{"denominator":66528,"numerator":443,"percentage_points":"+0.67 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_6d26a2eb1f68b5c6272c41202e6e75991396d01079402fb6242dd14dad3d1390","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_7289d0bfa90f78c9d37e3c289ff16a5e5d4335b72db4b1f0ba5fd894f405989c","model_configuration":"gpt-4-0613","observed_competition_rank":225,"organization":"OpenAI","original":{"denominator":3168,"numerator":971,"percent":"30.65%"},"original_question_n":198,"protocol":"mean@16","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":16,"clean":{"denominator":16,"numerator":5,"percent":"31.25%"},"clean_question_n":189,"configuration_date":"2025-01-21","configuration_id":"gpqa_cfg_v1_18fba063eae1be42dd8389b74872396a60463f5d291bf70c9532cbf3f5cd9e92","delta":{"denominator":1056,"numerator":7,"percentage_points":"+0.66 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_eec81bd37fd42fc5713cf1fcb43fa19910c58f9ff22360b4b92859007e4be14b","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_03c820e2b115ed14cf9d3a0b8a3ad2a1fa0e09efdfda594c4c7028fbb7d8b484","model_configuration":"Mixtral-8x7B-Instruct-v0.1","observed_competition_rank":226,"organization":"Mistral AI","original":{"denominator":1056,"numerator":323,"percent":"30.59%"},"original_question_n":198,"protocol":"mean@16","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":16,"clean":{"denominator":216,"numerator":67,"percent":"31.02%"},"clean_question_n":189,"configuration_date":"2025-01-20","configuration_id":"gpqa_cfg_v1_e20f7fe639d9555d87a238b746aca2117c1f00aca9069bf5ace08476f215d92f","delta":{"denominator":4752,"numerator":13,"percentage_points":"+0.27 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_87b9b646a70b83e9dee4e484887cb3697e1645d3bafca76160ca769c766338ca","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_9366d13835c1409504e6a6524428eaf482ae1c956f2dd8dd4d32a7fc49320512","model_configuration":"qwen1.5-32b-chat","observed_competition_rank":227,"organization":"Alibaba","original":{"denominator":1584,"numerator":487,"percent":"30.74%"},"original_question_n":198,"protocol":"mean@16","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":16,"clean":{"denominator":756,"numerator":229,"percent":"30.29%"},"clean_question_n":189,"configuration_date":"2025-01-22","configuration_id":"gpqa_cfg_v1_9ea8f6cc0477f44d90b6a00049e5e0d5848d4514109e686a2ebe2c8c74ade197","delta":{"denominator":66528,"numerator":307,"percentage_points":"+0.46 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_f43b71725dbadfc796496d181aa4ea130e427290e208c538c7f33456aaff2d61","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_e96691995d2919032340f96d1f25fa80b649eebb0736dbf4e4d03b3b6589517c","model_configuration":"open-mixtral-8x7b","observed_competition_rank":228,"organization":"Mistral AI","original":{"denominator":352,"numerator":105,"percent":"29.83%"},"original_question_n":198,"protocol":"mean@16","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":16,"clean":{"denominator":3024,"numerator":907,"percent":"29.99%"},"clean_question_n":189,"configuration_date":"2025-01-22","configuration_id":"gpqa_cfg_v1_39ebb391cd30cb3a52b963dd160804c308e182b68c53d4420593c1241547bbb7","delta":{"denominator":66528,"numerator":67,"percentage_points":"+0.10 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_cde3565a56b1a6656dafbec075f1438586402733203d291442929aa5a0fe8e98","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_d90fff655dcfbdf0cfe93c973d7a1ee242c06aee473853daa7a5b1fa26756c4e","model_configuration":"open-mistral-nemo-2407","observed_competition_rank":229,"organization":"Mistral AI","original":{"denominator":3168,"numerator":947,"percent":"29.89%"},"original_question_n":198,"protocol":"mean@16","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":16,"clean":{"denominator":756,"numerator":223,"percent":"29.50%"},"clean_question_n":189,"configuration_date":"2025-01-20","configuration_id":"gpqa_cfg_v1_bcfd3a2eaf724a8a23244b602b5694e0a00e463c8d82eb1c2c4ab4333abe3a1e","delta":{"denominator":6048,"numerator":41,"percentage_points":"+0.68 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_5f85db9f6e9ea1d09922a756afb331f3156ee27cd024227e1bfa68f46e145174","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_98d4be1393e2c39c14ec82adf828f8420be9fc3d887c76cb23379312fa4b18a2","model_configuration":"qwen1.5-72b-chat","observed_competition_rank":230,"organization":"Alibaba","original":{"denominator":288,"numerator":83,"percent":"28.82%"},"original_question_n":198,"protocol":"mean@16","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":16,"clean":{"denominator":1008,"numerator":281,"percent":"27.88%"},"clean_question_n":189,"configuration_date":"2025-01-17","configuration_id":"gpqa_cfg_v1_08bd8befe0cade55c7342c9f96b69d027677412b414b670247ac45b3f919f540","delta":{"denominator":11088,"numerator":-17,"percentage_points":"-0.15 pp"},"delta_direction":"decrease","evaluation_run_id":"gpqa_run_v1_5e39fd30f2fea819992cae499e5283ef32d1247ded1e8b67c327c11717274ea5","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_19777c5b3118044034b1d9244e71cd858bbdb22012be1097744b6064e7d9b84e","model_configuration":"gpt-3.5-turbo-1106","observed_competition_rank":231,"organization":"OpenAI","original":{"denominator":132,"numerator":37,"percent":"28.03%"},"original_question_n":198,"protocol":"mean@16","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":16,"clean":{"denominator":504,"numerator":139,"percent":"27.58%"},"clean_question_n":189,"configuration_date":"2025-01-31","configuration_id":"gpqa_cfg_v1_79ea3c485cc923af3af9258fe5fa39ad7844f3c46260f2a10709c1ff9e1b57e6","delta":{"denominator":11088,"numerator":-1,"percentage_points":"-0.01 pp"},"delta_direction":"decrease","evaluation_run_id":"gpqa_run_v1_e5ae10b7fa25360b7eab7f9fbf1ab5109372ab3b9e4f3b062ad7eead22f752dd","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_99902969eaf1717bb1a5557a5e692c80987156860a36229e2c160b0f37eb0d17","model_configuration":"Phi-3-medium-128k-instruct","observed_competition_rank":232,"organization":"Microsoft","original":{"denominator":1584,"numerator":437,"percent":"27.59%"},"original_question_n":198,"protocol":"mean@16","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":16,"clean":{"denominator":504,"numerator":139,"percent":"27.58%"},"clean_question_n":189,"configuration_date":"2025-01-17","configuration_id":"gpqa_cfg_v1_c486088d5d896e439d15de33eede35061d90e78929121ccb49dcd5c09ad7b1cd","delta":{"denominator":11088,"numerator":13,"percentage_points":"+0.12 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_88afd62f888feaa64bfd19ce1b3107b7a12ceaf768680d041bc5cbf374a06b80","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_52c8fadd084b1897a73631b39ea57ce0a8ffbec358a2975f7a95906190039081","model_configuration":"gemma-2-9b-it","observed_competition_rank":232,"organization":"Google DeepMind","original":{"denominator":528,"numerator":145,"percent":"27.46%"},"original_question_n":198,"protocol":"mean@16","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":16,"clean":{"denominator":216,"numerator":59,"percent":"27.31%"},"clean_question_n":189,"configuration_date":"2025-01-17","configuration_id":"gpqa_cfg_v1_4ff2a56599222d2a455b2c7961cc7f544af144e423b64cfd6a02e8ed9c88e446","delta":{"denominator":9504,"numerator":13,"percentage_points":"+0.14 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_6f145229001723016030c8489246da79111c7919d38127a356e8a0ad74a3ca45","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_c3994ed498ac294c3fb7ad1c0b4d15590978b729fabd6bf692874c979d1ecccb","model_configuration":"gpt-3.5-turbo-0125","observed_competition_rank":234,"organization":"OpenAI","original":{"denominator":1056,"numerator":287,"percent":"27.18%"},"original_question_n":198,"protocol":"mean@16","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":16,"clean":{"denominator":48,"numerator":13,"percent":"27.08%"},"clean_question_n":189,"configuration_date":"2025-01-21","configuration_id":"gpqa_cfg_v1_48c8d1776d8586ba5fce19ed21b2148cc8f42ab5c360ce1f53012f8973abf633","delta":{"denominator":132,"numerator":1,"percentage_points":"+0.76 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_711746500282f901b2cd94eb120ad984ae3a3f95786e881c2153bbfa1889a4ae","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_31aec9986e3f2401cfd7f1131e8c104af9961fb562e2fef154db8a1e512e2f92","model_configuration":"Llama-2-70b-chat-hf","observed_competition_rank":235,"organization":"Meta AI","original":{"denominator":528,"numerator":139,"percent":"26.33%"},"original_question_n":198,"protocol":"mean@16","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":16,"clean":{"denominator":189,"numerator":50,"percent":"26.46%"},"clean_question_n":189,"configuration_date":"2025-01-22","configuration_id":"gpqa_cfg_v1_08f66b9ca9530d2a229621c663831d6ac6cfa888b11230239eb991363750e754","delta":{"denominator":16632,"numerator":-115,"percentage_points":"-0.69 pp"},"delta_direction":"decrease","evaluation_run_id":"gpqa_run_v1_dbb9f2125cc09e996ef57fb0c3a86c5039790a3328752ace68ac73c52fd858b8","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_4c3fc342c1502e5abf9c4a73c86e03d99ee9af7721e0cac0d97df58e9b65a1fc","model_configuration":"ministral-8b-2410","observed_competition_rank":236,"organization":"Mistral AI","original":{"denominator":792,"numerator":215,"percent":"27.15%"},"original_question_n":198,"protocol":"mean@16","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":16,"clean":{"denominator":3024,"numerator":797,"percent":"26.36%"},"clean_question_n":189,"configuration_date":"2025-01-21","configuration_id":"gpqa_cfg_v1_3166c2961eb47a41245ec120bbd6aa00af424fb8d90af97fb2168f58e13fb285","delta":{"denominator":16632,"numerator":47,"percentage_points":"+0.28 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_a3f20303482242d26463b4d6d08959c23efb1301c1a7e6f91bd759ea7f96aae5","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_27e29c8b2a4642558e73c768888c044653ecc7516793e83a566599713822dc17","model_configuration":"Meta-Llama-3-8B-Instruct","observed_competition_rank":237,"organization":"Meta AI","original":{"denominator":1584,"numerator":413,"percent":"26.07%"},"original_question_n":198,"protocol":"mean@16","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":16,"clean":{"denominator":3024,"numerator":793,"percent":"26.22%"},"clean_question_n":189,"configuration_date":"2025-01-20","configuration_id":"gpqa_cfg_v1_b6ca95c2ac417673b46a71c78ad3dd16d60626c176c639d56b530e84b8805233","delta":{"denominator":8316,"numerator":23,"percentage_points":"+0.28 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_fe427bef732f5f1a516b2590fa00be675554c2e40a20fa0ead1cc7602785c293","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_56b4fdcabe7d5a6922b99b83db483de7ec02e83ebc8cffacfb60a3b0382b154e","model_configuration":"Llama-3.1-8B-Instruct","observed_competition_rank":238,"organization":"Meta AI","original":{"denominator":528,"numerator":137,"percent":"25.95%"},"original_question_n":198,"protocol":"mean@16","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":16,"clean":{"denominator":126,"numerator":31,"percent":"24.60%"},"clean_question_n":189,"configuration_date":"2025-01-22","configuration_id":"gpqa_cfg_v1_139cbaca23c931c65e3d28182449928faf23acb7ce5481a3b9bb7577edf379de","delta":{"denominator":154,"numerator":-1,"percentage_points":"-0.65 pp"},"delta_direction":"decrease","evaluation_run_id":"gpqa_run_v1_8feb80681a8b5c93af2d6f169a81b92d5277ab2284b7c0eff8bf8507f840f244","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_eefbb4d726b36d38e57429d4802a9358be6e10e1c00ab79f8cbae4ccdf8e81ea","model_configuration":"ministral-3b-2410","observed_competition_rank":239,"organization":"Mistral AI","original":{"denominator":99,"numerator":25,"percent":"25.25%"},"original_question_n":198,"protocol":"mean@16","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":16,"clean":{"denominator":216,"numerator":53,"percent":"24.54%"},"clean_question_n":189,"configuration_date":"2025-01-21","configuration_id":"gpqa_cfg_v1_937900a08828620bb405325f43551909f02e0fa51fe3dcfd74cfe0f34f2404f2","delta":{"denominator":1188,"numerator":-1,"percentage_points":"-0.08 pp"},"delta_direction":"decrease","evaluation_run_id":"gpqa_run_v1_c6bd22f4f5247b1f051af18d19da4bdc9c1f2d209d93da1064e0b851bfd1fd34","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_a824ee6311547ec13d7bd99dbe7f19fad0f95c1046e582fa09887b8f15ef5e6b","model_configuration":"deepseek-llm-67b-chat","observed_competition_rank":240,"organization":"DeepSeek","original":{"denominator":264,"numerator":65,"percent":"24.62%"},"original_question_n":198,"protocol":"mean@16","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":16,"clean":{"denominator":3024,"numerator":467,"percent":"15.44%"},"clean_question_n":189,"configuration_date":"2025-01-21","configuration_id":"gpqa_cfg_v1_7f2ac119b9c9f420c06ad2e6f93c6e2e82828c27cec46dd73558c7faf28e5573","delta":{"denominator":66528,"numerator":173,"percentage_points":"+0.26 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_6fdcfcbc939cac11581b5cd167f87ce8ad1a8c643783737ff83f0d36485b9eec","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_908d2fb385d65caf4e9235b3a9d0cadd7f65a37115bc964222f5247ad188dadc","model_configuration":"Mistral-7B-Instruct-v0.3","observed_competition_rank":241,"organization":"Mistral AI","original":{"denominator":3168,"numerator":481,"percent":"15.18%"},"original_question_n":198,"protocol":"mean@16","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":16,"clean":{"denominator":756,"numerator":115,"percent":"15.21%"},"clean_question_n":189,"configuration_date":"2025-01-20","configuration_id":"gpqa_cfg_v1_5e3089906c8c2437abc826f155bc422a429bb1076e8eaaa770f17c305e3498ab","delta":{"denominator":66528,"numerator":313,"percentage_points":"+0.47 pp"},"delta_direction":"increase","evaluation_run_id":"gpqa_run_v1_cc3f60648c996eaaa52ecbef11d0122f18e87ca4999935be5a754209ac741c6d","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_fae0271f6a0c6911f2f54f368c94835199df8d329517cae7cc978051576c7b30","model_configuration":"Yi-34B-Chat","observed_competition_rank":242,"organization":"01.AI","original":{"denominator":3168,"numerator":467,"percent":"14.74%"},"original_question_n":198,"protocol":"mean@16","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"},{"attempts_per_question":16,"clean":{"denominator":84,"numerator":11,"percent":"13.10%"},"clean_question_n":189,"configuration_date":"2025-01-22","configuration_id":"gpqa_cfg_v1_00e9ae21c6c0f764c862027370ab0a348eb54d424a8aff0cb6a305a38618706b","delta":{"denominator":22176,"numerator":-29,"percentage_points":"-0.13 pp"},"delta_direction":"decrease","evaluation_run_id":"gpqa_run_v1_ca45a83a12af45522007a22520a874ec87c40ccee3a08cb3975b2f21710f3022","evidence_class":"direct_compatibility","evidence_variant_id":"gpqa_evd_v1_b1fbd626729590cba9a7306c14eab42a7a0746d386cef5fb1bc3ed7c5ff05c7e","model_configuration":"open-mistral-7b","observed_competition_rank":243,"organization":"Mistral AI","original":{"denominator":3168,"numerator":419,"percent":"13.23%"},"original_question_n":198,"protocol":"mean@16","release_id":"gpqa-clean-results-v0.1.0","schema_version":"gpqa-results-public-leaderboard-v2","validation_profile":"compatibility_complete_legacy_header"}],"schema_version":"gpqa-clean-results-leaderboard-v2","summary":{"configuration_count":243,"evidence_class_counts":{"derived_uniform":6,"direct_compatibility":116,"direct_strict":108,"project_native":9,"targeted_recovery":4},"protocol_counts":{"mean@1":116,"mean@16":63,"mean@2":2,"mean@6":1,"mean@7":4,"mean@8":57}}}
