# CritPt

Unpublished research-level physics problems written by working physicists, in areas like quantum physics and astrophysics. It tests whether AI can do the kind of reasoning needed in real scientific research.

Category: reasoning. Type: dataset. Status: active.
Metric: Share of problems with the correct final answer (%) (higher is better).
Source: Epoch AI (Benchmarking Hub) — https://epoch.ai/data/benchmark_data.zip, results as of 2026-11-17, retrieved 2026-11-17, license: CC-BY-4.0.

| Rank | Model | Organization | Accuracy (%) | Release date |
|---|---|---|---|---|
| 1 | Gemini 4 Carbon (High) | Google DeepMind | 36.0 | 2026-11-10 |
| 2 | Claude Fable 5.2 (Xhigh, Default Fallback) | Anthropic | 35.7 | 2026-11-09 |
| 3 | Claude Fable 5.2 (Max, Default Fallback) | Anthropic | 35.4 | 2026-11-09 |
| 4 | Claude Fable 5.2 (High, Default Fallback) | Anthropic | 32.9 | 2026-11-09 |
| 5 | GPT-5.6 Sol (Max) | OpenAI | 32.3 | 2026-07-09 |
| 6 | GPT-6 Astra (Max) | OpenAI | 31.7 | 2026-09-03 |
| 7 | Claude Opus 5.5 (Max, Default Fallback) | Anthropic | 31.7 | 2026-09-22 |
| 8 | Claude Opus 5.5 (Xhigh, Default Fallback) | Anthropic | 31.7 | 2026-09-22 |
| 9 | GPT-6.1 Sol (Max) | OpenAI | 31.7 | 2026-09-29 |
| 10 | GPT-6.1 Sol (Xhigh) | OpenAI | 31.7 | 2026-09-29 |
| 11 | GPT-6 Astra (Xhigh) | OpenAI | 31.4 | 2026-09-03 |
| 12 | Claude Sonnet 5.5 (Max, Default Fallback) | Anthropic | 31.4 | 2026-09-28 |
| 13 | Claude Fable 5.1 (Xhigh, Default Fallback) | Anthropic | 31.1 | 2026-09-01 |
| 14 | Claude Sonnet 5.5 (Xhigh, Default Fallback) | Anthropic | 31.1 | 2026-09-28 |
| 15 | GPT-6 Sol (Max) | OpenAI | 30.9 | 2026-09-22 |
| 16 | Claude Opus 5.5 (High, Default Fallback) | Anthropic | 30.9 | 2026-09-22 |
| 17 | Claude Fable 5.2 (Medium, Default Fallback) | Anthropic | 30.9 | 2026-11-09 |
| 18 | GPT-5.5 Pro (Xhigh) | OpenAI | 30.6 | 2026-04-23 |
| 19 | Claude Fable 5.1 (High, Default Fallback) | Anthropic | 30.3 | 2026-09-01 |
| 20 | GPT-5.4 Pro (Xhigh) | OpenAI | 30.0 | 2026-03-05 |
| 21 | GPT-5.6 Terra (Max) | OpenAI | 30.0 | 2026-07-09 |
| 22 | GPT-6.1 Sol (High) | OpenAI | 30.0 | 2026-09-29 |
| 23 | Claude Fable 5.1 (Max, Default Fallback) | Anthropic | 29.7 | 2026-09-01 |
| 24 | Claude Fable 5.2 (Low, Default Fallback) | Anthropic | 29.4 | 2026-11-09 |
| 25 | Claude Opus 5 (Adaptive Reasoning, Max Effort) | Anthropic | 29.1 | 2026-07-24 |

Notes: External results mirrored by Epoch AI (critpt_external.csv), converted to % (1 decimal). One row per model + reasoning-effort configuration.
