METR Time Horizons
Each model’s result as Epoch AI’s Benchmarking Hub publishes it, run by Epoch or collected from the source it cites. Facthory has not run them.
Data: Epoch AI, ‘Capabilities & benchmarking’, epoch.ai/benchmarks, licensed CC BY 4.0. Read on 16 Nov 2026.
The chart shows the 25 best of 46 results; the table lists them all.
| # | Model | Organization | Result | Evidence |
|---|---|---|---|---|
| 1 | Claude Mythos Preview (Early) Limited access | Anthropic | 85.2% | – |
| 2 | Claude Opus 4.6 (unknown thinking) API access | Anthropic | 78.9% | – |
| 3 | Gemini 3.1 Pro Preview API access | Google DeepMind | 77% | – |
| 4 | GPT-5.2 (high) API access | OpenAI | 75.3% | – |
| 5 | Claude Opus 4.5 (16k thinking) API access | Anthropic | 75% | Log or source |
| 6 | GPT-5.3 Codex API access | OpenAI | 74.5% | – |
| 7 | GPT-5.4 (xhigh) API access | OpenAI | 74.3% | – |
| 8 | Claude Opus 4.5 (no thinking) API access | Anthropic | 73% | – |
| 9 | Gemini 3 Pro Preview API access | Google DeepMind | 71% | – |
| 10 | GPT-5.1-Codex-Max API access | OpenAI | 70.8% | Log or source |
| 11 | GPT-5 (medium) API access | OpenAI | 69.6% | Log or source |
| 12 | GPT-5 (high) API access | OpenAI | 69.4% | – |
| 13 | Claude Sonnet 4.5 (16k thinking) API access | Anthropic | 67.4% | Log or source |
| 14 | claude-opus-4-1-20250805_16K API access | Anthropic | 66.8% | Log or source |
| 15 | grok-4-0709 API access | xAI | 66.6% | Log or source |
| 16 | o3 (medium) API access | OpenAI | 65.4% | Log or source |
| 17 | Claude Opus 4 API access | Anthropic | 63.9% | Log or source |
| 18 | o4-mini (medium) API access | OpenAI | 63.9% | Log or source |
| 19 | o3 (unknown thinking) API access | OpenAI | 63.6% | – |
| 20 | claude-sonnet-4-20250514_16K API access | Anthropic | 62% | Log or source |
| 21 | claude-opus-4-1-20250805 API access | Anthropic | 61.6% | – |
| 22 | Claude Opus 4 API access | Anthropic | 61.5% | – |
| 23 | Claude 3.7 Sonnet (16k thinking) API access | Anthropic | 60% | Log or source |
| 24 | Kimi K2 Thinking Open weights (restricted use) | Moonshot | 59.2% | Log or source |
| 25 | gpt-oss-120b Open weights (unrestricted) | OpenAI | 56.6% | Log or source |
| 26 | claude-3-7-sonnet-20250219 API access | Anthropic | 55.8% | – |
| 27 | Gemini 2.5 Pro Preview (Jun 2025) API access | Google DeepMind | 55.4% | Log or source |
| 28 | DeepSeek-R1 (May 2025) Open weights (unrestricted) | DeepSeek | 53.8% | Log or source |
| 29 | DeepSeek-R1 Open weights (unrestricted) | DeepSeek | 51.9% | Log or source |
| 30 | o1 (medium) API access | OpenAI | 51.1% | Log or source |
| 31 | DeepSeek-V3 (Mar 2025) Open weights (restricted use) | DeepSeek | 49.6% | Log or source |
| 32 | DeepSeek-V3 Open weights (restricted use) | DeepSeek | 47.4% | Log or source |
| 33 | Claude 3.5 Sonnet (Oct 2024) API access | Anthropic | 45.2% | Log or source |
| 34 | o1-preview API access | OpenAI | 45.1% | Log or source |
| 35 | GPT-4o (May 2024) API access | OpenAI | 40.8% | Log or source |
| 36 | Claude 3.5 Sonnet (Jun 2024) API access | Anthropic | 40.2% | Log or source |
| 37 | gpt-4-turbo-2024-04-09 API access | OpenAI | 36.7% | Log or source |
| 38 | GPT-4 (Mar 2023) API access | OpenAI | 36.1% | Log or source |
| 39 | qwen2.5-72b-instruct Open weights (unrestricted) | Alibaba | 35.8% | Log or source |
| 40 | GPT-4 Turbo Preview (January 2024) API access | OpenAI | 35.2% | Log or source |
| 41 | qwen2-72b-instruct Open weights (unrestricted) | Alibaba | 29.9% | Log or source |
| 42 | claude-3-opus-20240229 API access | Anthropic | 29.5% | Log or source |
| 43 | GPT-4 Turbo Preview (Nov 2023) API access | OpenAI | 28.9% | Log or source |
| 44 | gpt-3.5-turbo-instruct API access | OpenAI | 21.5% | Log or source |
| 45 | davinci-002 API access | OpenAI | 16.2% | Log or source |
| 46 | gpt2-xl Open weights (unrestricted) | OpenAI | 10.1% | Log or source |
