Skip to content

OpenAI

GPT-5

Released with mini and nano versions in the API the same day. METR time horizon: about 3.4 hours.

Announced
Access
Public

Benchmark scores

TestScoreBest on the trackerHow it was measured
Epoch Capabilities Indexmany tests combined into one score150 pointsGPT-6.2 Astra 174 pointsIndependentlisted as "GPT-5", scale anchor (no CI)Run 9 Nov 2026
Remote Labor Indexpaid freelance projects1.7%Claude Fable 5.2 41%Independentgpt-5-2025-08-07Run 27 Oct 2025
MirrorCoderebuilding whole programs4.4%Claude Opus 5.5 79%Independenthigh reasoning effort, MirrorCode (ML, +Private, 2L): 15 programs x 2 languages, 3 runs each, 10B-token / 7-day budget, ±3.4 SERun 21 Oct 2026
Humanity's Last Examexpert questions, no tools25%GPT-6 Astra 55%Independentleaderboard entry "gpt-5-2025-08-07" (gpt-5-2025-08-07_high)
FrontierMath T1–3hard maths problems55%Claude Fable 5.2 95%Independenthigh reasoning effort (gpt-5-2025-08-07_high)Run 10 Jun 2026
GPQA DiamondPhD-level science questionsPhD experts: 70%Saturated86%Gemini 4 Argon 96%Independenthigh reasoning effort (gpt-5-2025-08-07_high)Run 29 Oct 2025
FrontierMath T4research-level maths problemsSaturated22%GPT-6.1 Sol 100%Independenthigh reasoning effort (gpt-5-2025-08-07_high)Run 11 Jun 2026
ARC-AGI-2visual puzzles, model aloneAverage person: 60%Saturated9.9%GPT-6 Astra 95%IndependentARC Prize config "GPT-5 (High)" (gpt-5-2025-08-07-high)
OTIS Mock AIMEcompetition mathsSaturated91%GPT-5.5 100%Independenthigh reasoning effort (gpt-5-2025-08-07_high)Run 29 Oct 2025
SWE-bench Verifiedfixing real software bugsKnown flaws74%Claude Opus 4.7 83%Independenthigh reasoning effort (gpt-5-2025-08-07_high)Run 6 Feb 2026
METR time horizon (TH1.1 + TH2)task length an AI agent finishes half the time3 hGPT-6 Astra 1.7 daysIndependentPooled TH1.1 + TH2 suites; 95% CI 97.9–334.5 min
METR time horizontask length an AI agent finishes half the timeNo longer updated3.4 hClaude Mythos Preview 17 hIndependentMETR Horizon v1.1 run 'gpt_5_2025_08_07_inspect', high reasoning effort (per Epoch AI mapping); 95% CI 112.6–405.6 min

Compare every model on the same tests →

Also on these charts

Sources