OmniPhys - Mechanics - Answer Accuracy: leaderboard
OmniPhys (arXiv 2608.25398v1, Table 3): multimodal physics problems from junior high school to university, mostly from Chinese educational materials, for 26 MLLMs; process scores average DeepSeek-V3.2 and GPT-4 judges. S1 result score on objective mechanics tasks; percent, higher is better.
Metric: Result score (%; objective tasks, S1). Source: arxiv.org. Saturation forecast: Around February 2027. 26 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3 Pro | 83.63 | #82 |
| 2 | Seed-1.6 | 82.64 | #272 |
| 3 | GLM-4.6V | 75.05 | #330 |
| 4 | Qwen 3 VL 235B A22B | 70.72 | #251 |
| 5 | Gemini 2.5 Flash | 69.5 | #255 |
| 6 | GPT-5.2 | 66.33 | #115 |
| 7 | O4 Mini | 64.85 | #185 |
| 8 | Claude Sonnet 4.5 | 54.24 | #148 |
| 9 | GPT-5.1 | 52.3 | #145 |
| 10 | Grok 4 | 47.91 | #189 |
| 11 | Qwen 2.5 VL 7B | 39.72 | #641 |
| 12 | GPT-4o | 39.41 | #357 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=omniphys-mechanics-answer-accuracy · How It Works · Data refreshed daily, snapshot 2026-11-05.