Physical Perception on PhysUniBench
75PhysUniBench ScoreClaude-4.5-Haiku
Evaluation Results
| Method | Links | |
|---|---|---|
| Claude-4.5-HaikuSize=-2026.02 | 75 | |
| GPT-5Size=-2026.02 | 71.9 | |
| Gemini-2.5-flashSize=-2026.02 | 65.5 | |
| OmniFysicsSize=3B2026.02 | 50.8 | |
| Qwen3-VL-8B-InstructSize=8B2026.02 | 44.6 | |
| InternVL3.5-8BSize=8B2026.02 | 43.8 | |
| SAIL-VL2Size=2B2026.02 | 37.3 | |
| Ovis2.5Size=2B2026.02 | 37 | |
| Qwen2.5-OmniSize=3B2026.02 | 33.4 |