OpenAI
GPT-5.2
Aimed at professional work, released with GPT-5.2 Pro. METR time horizon: about 5.9 hours.
- Announced
- Access
- Public
Benchmark scores
| Test | Score | Best on the tracker | How it was measured |
|---|---|---|---|
| Epoch Capabilities Indexmany tests combined into one score | 154 points | GPT-6.2 Astra 174 points | Independentlisted as "GPT-5.2", CI 152.77 to 156.24Run 9 Nov 2026 |
| Remote Labor Indexpaid freelance projects | 2.5% | Claude Fable 5.2 41% | Independentgpt-5.2-2025-12-11 (medium)Run 29 Jan 2026 |
| Humanity's Last Examexpert questions, no tools | 28% | GPT-6 Astra 55% | Independentleaderboard entry "gpt-5.2-2025-12-11" (gpt-5.2-2025-12-11_unknown) |
| FrontierMath T1–3hard maths problems | 67% | Claude Fable 5.2 95% | Independentxhigh reasoning effort (gpt-5.2-2025-12-11_xhigh)Run 11 Jun 2026 |
| GPQA DiamondPhD-level science questionsPhD experts: 70%Saturated | 91% | Gemini 4 Argon 96% | Independentxhigh reasoning effort (gpt-5.2-2025-12-11_xhigh)Run 13 Dec 2025 |
| FrontierMath T4research-level maths problemsSaturated | 32% | GPT-6.1 Sol 100% | Independentxhigh reasoning effort (gpt-5.2-2025-12-11_xhigh)Run 11 Jun 2026 |
| ARC-AGI-2visual puzzles, model aloneAverage person: 60%Saturated | 53% | GPT-6 Astra 95% | IndependentARC Prize config "GPT-5.2 (XHigh)" (gpt-5-2-2025-12-11-thinking-xhigh) |
| OTIS Mock AIMEcompetition mathsSaturated | 96% | GPT-5.5 100% | Independenthigh reasoning effort (gpt-5.2-2025-12-11_high)Run 11 Dec 2025 |
| SWE-bench Verifiedfixing real software bugsKnown flaws | 74% | Claude Opus 4.7 83% | Independenthigh reasoning effort (gpt-5.2-2025-12-11_high)Run 12 Feb 2026 |
| METR time horizontask length an AI agent finishes half the timeNo longer updated | 5.9 h | Claude Mythos Preview 17 h | IndependentMETR Horizon v1.1 run 'gpt_5_2', high reasoning effort (per Epoch AI mapping); 95% CI 198.1–815.2 min |
Also on these charts
Sources
- Introducing GPT-5.2OpenAI · Official · openai.com
- OpenAI API changelogOpenAI · Official · developers.openai.com
- METR time horizonsMETR · Official · metr.org
- Epoch AI Benchmarking Hub (model release dates and runs)Epoch AI · Dataset · 9 Nov 2026 · epoch.ai