Home Model series GPT unsuffixed models GPT unsuffixed models · Humanity's Last Exam · 2,500-question benchmark · Release history GPT unsuffixed models · Humanity's Last Exam · 2,500-question benchmark · Release history View scores, configurations, and disclosures for Humanity's Last Exam · 2,500-question benchmark across 4 GPT unsuffixed models releases, ordered by release date.
Differences are calculated only for unique, active, exact values between direct predecessors when the unit, metric, and explicitly recorded configuration and conditions match.
2025-08-07 · 1 score record for this release, including 1 exact value (may include historical records)
GPT-5 · Humanity's Last Exam scores and configurations Metric and score Configuration and evaluation conditions Sources Accuracy 24.8% Reported 24.8% · Unit % Published 2025-08-07
GPT-5 (high) Settings API reasoning model; the developer release benchmark table labels high reasoning effort. Unknowns none recorded Tools Tools forbidden Reasoning budget high Framework not recorded Protocol HLE (no tools) Conditions API GPT-5; high reasoning effort. View this score in rankings
Detailed benchmarks / HLE (no tools) Main text · Introducing GPT-5 for developers | OpenAI 2025-12-11 · 4 score records for this release, including 4 exact values (may include historical records)
GPT-5.2 · Humanity's Last Exam scores and configurations Metric and score Configuration and evaluation conditions Sources Accuracy 34.5% Reported 34.5% · Unit % Published Unconfirmed
GPT-5.2 Thinking (xhigh) Settings Highest available API reasoning effort xhigh, except professional benchmark tasks. Unknowns none recorded Tools Tools forbidden Reasoning budget xhigh Framework not recorded Protocol HLE, no tools; exact subset/version not restated in table. Conditions Highest available API reasoning effort xhigh. View this score in rankings Accuracy 36.6% Reported 36.6% · Unit % Published Unconfirmed
GPT-5.2 Pro (xhigh) Settings Pro variant; highest available API reasoning effort xhigh except professional benchmark tasks. Unknowns The article does not publish a Pro GDPval result under the same heavy setting as the Thinking result; reported row retained as printed. Tools Tools forbidden Reasoning budget xhigh Framework not recorded Protocol HLE, no tools; exact subset/version not restated in table. Conditions GPT-5.2 Pro; highest available API effort xhigh. View this score in rankings Accuracy 45.5% Reported 45.5% · Unit % Published Unconfirmed
GPT-5.2 Thinking (xhigh) Settings Highest available API reasoning effort xhigh, except professional benchmark tasks. Unknowns none recorded Tools Tools allowed Reasoning budget xhigh Framework not recorded Protocol HLE with search and Python; exact subset/version not restated in table. Conditions Highest available API reasoning effort xhigh. View this score in rankings Accuracy 50% Reported 50.0% · Unit % Published Unconfirmed
GPT-5.2 Pro (xhigh) Settings Pro variant; highest available API reasoning effort xhigh except professional benchmark tasks. Unknowns The article does not publish a Pro GDPval result under the same heavy setting as the Thinking result; reported row retained as printed. Tools Tools allowed Reasoning budget xhigh Framework not recorded Protocol HLE with search and Python; exact subset/version not restated in table. Conditions GPT-5.2 Pro; highest available API effort xhigh. View this score in rankings
Appendix / Academic / HLE rows Main text · 隆重推出 GPT-5.2 | OpenAI 2026-03-05 · 4 score records for this release, including 4 exact values (may include historical records)
Records for the direct predecessor do not provide matching conditions for a valid difference.
GPT-5.4 · Humanity's Last Exam scores and configurations Metric and score Configuration and evaluation conditions Sources Accuracy 39.8% Reported 39.8% · Unit % Published Unconfirmed
GPT-5.4 (xhigh) Settings The release article states evaluations use xhigh unless otherwise noted. Unknowns none recorded Tools Tools forbidden Reasoning budget xhigh Framework not recorded Protocol Humanity's Last Exam; no tools. Conditions Official release table labels GPT-5.4 (xhigh); the article says evaluations use xhigh unless otherwise noted and were run in a research environment. View this score in rankings Accuracy 42.7% Reported 42.7% · Unit % Published Unconfirmed
GPT-5.4 Pro Settings Pro model variant; the article reports Pro separately from GPT-5.4. Unknowns The Pro column is not a reasoning-effort label; the exact effort for each Pro evaluation is not separately stated. Tools Tools forbidden Reasoning budget not recorded Framework not recorded Protocol Humanity's Last Exam; no tools. Conditions Official release table labels GPT-5.4 Pro; research evaluation environment. View this score in rankings Accuracy 52.1% Reported 52.1% · Unit % Published Unconfirmed
GPT-5.4 (xhigh) Settings The release article states evaluations use xhigh unless otherwise noted. Unknowns none recorded Tools Tools allowed Reasoning budget xhigh Framework not recorded Protocol Humanity's Last Exam; with tools. Conditions Official release table labels GPT-5.4 (xhigh); the article says evaluations use xhigh unless otherwise noted and were run in a research environment. View this score in rankings Accuracy 58.7% Reported 58.7% · Unit % Published Unconfirmed
GPT-5.4 Pro Settings Pro model variant; the article reports Pro separately from GPT-5.4. Unknowns The Pro column is not a reasoning-effort label; the exact effort for each Pro evaluation is not separately stated. Tools Tools allowed Reasoning budget not recorded Framework not recorded Protocol Humanity's Last Exam; with tools. Conditions Official release table labels GPT-5.4 Pro; research evaluation environment. View this score in rankings
2026-04-23 · 2 score records for this release, including 2 exact values (may include historical records)
GPT-5.5 · Humanity's Last Exam scores and configurations Metric and score Configuration and evaluation conditions Sources Accuracy 41.4% Reported 41.4% · Unit % Published Unconfirmed
GPT-5.5 xhigh Settings Reasoning effort xhigh; research environment; output may differ slightly from production ChatGPT in some cases. Unknowns A subset of protocol details remain benchmark-specific and unreported. Tools Tools forbidden Reasoning budget xhigh Framework not recorded Protocol not specified Conditions Mainline GPT-5.5 evaluation at xhigh in research environment. View this score in rankings Accuracy 43.1% Reported 43.1% · Unit % Published Unconfirmed
GPT-5.5 Pro Settings Same underlying model with a setting that uses parallel test-time compute. Unknowns Parallel compute budget and exact Pro harness settings are not stated. Tools Tools forbidden Reasoning budget not recorded Framework not recorded Protocol not specified Conditions GPT-5.5 Pro; parallel test-time compute; exact compute budget unspecified. View this score in rankings