Physical Perception on PhysBench
64.4PhysBench ScoreGPT-5
Evaluation Results
| Method | Links | |
|---|---|---|
| GPT-5Size=-2026.02 | 64.4 | |
| Gemini-2.5-flashSize=-2026.02 | 56.4 | |
| Qwen3-VL-8B-InstructSize=8B2026.02 | 52.1 | |
| OmniFysicsSize=3B2026.02 | 47.2 | |
| Claude-4.5-HaikuSize=-2026.02 | 45.4 | |
| SAIL-VL2Size=2B2026.02 | 44.4 | |
| Ovis2.5Size=2B2026.02 | 43.8 | |
| Qwen2.5-OmniSize=3B2026.02 | 35.5 | |
| InternVL3.5-8BSize=8B2026.02 | 31.4 |