PhysicsFinals: leaderboard
Benchmarklist's label for PHYSICS (Yale, NYU, Notre Dame, 2025): 1,297 PhD-qualifying-exam physics problems in six areas from mechanics to optics; free-form answers checked by SymPy then GPT-4o.
Metric: Score (imported). Source: arxiv.org. Source: benchmarklist.com. Saturation forecast: Estimated already saturated. 33 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | O3 Mini | 59.9 | #272 |
| 2 | DeepSeek R1 | 44.3 | #251 |
| 3 | O1 Mini | 43.6 | #323 |
| 4 | Gemini 1.5 Pro | 38.4 | #445 |
| 5 | GPT-4o | 36.7 | #343 |
| 6 | Claude 3.5 Sonnet | 34.7 | #350 |
| 7 | Qwen2.5-Math-72B | 32.2 | #727 |
| 8 | Llama 3.3 70B | 31.5 | #579 |
| 9 | Phi-4 | 29.1 | #706 |
| 10 | Qwen 2.5 72B | 28.7 | #499 |
| 11 | Qwen 2.5 32B | 27.6 | #546 |
| 12 | Qwen 2.5 7B | 20.4 | #922 |
| 13 | Qwen 2.5 14B | 19.6 | #725 |
| 14 | Gemma 2 27B | 18.3 | #806 |
| 15 | Yi 1.5 34B | 17.4 | #807 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=physicsfinals · How It Works · Data refreshed daily, snapshot 2026-10-20.