Spatial Reasoning on SAT
83.3Overall AccGPT-5
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| GPT-5Params=-2026.06 | 83.3 | — | — | — | — | — | |
| Qwen3-VL-InstructModel Size=235B-A22B, Type=Instruct2025.12 | 80 | 83.78 | 100 | 88.24 | 71.74 | 59.09 | |
| Gemini-2.5-ProType=Proprietary2025.12 | 79 | 85.14 | 100 | 77.94 | 73.91 | 62.12 | |
| CroPondModel Size=3B2025.12 | 78.33 | 82.43 | 100 | 92.65 | 58.7 | 57.58 | |
| RoboPINParams=4B2026.06 | 78 | — | — | — | — | — | |
| Gemini-2.5-ProParams=-2026.06 | 76.7 | — | — | — | — | — | |
| Embodied-R1Params=3B2026.06 | 76.3 | — | — | — | — | — | |
| RoboBrain2.0Params=7B2026.06 | 75.3 | — | — | — | — | — | |
| CroPondModel Size=7B2025.12 | 73 | 58.11 | 100 | 89.71 | 65.22 | 59.09 | |
| Qwen3-VLParams=4B2026.06 | 70.7 | — | — | — | — | — | |
| Qwen3-VL-InstructModel Size=30B-A3B, Type=Instruct2025.12 | 69.39 | 70.27 | 93.18 | 80.88 | 64.29 | 43.94 | |
| Qwen2.5-VL-InstructModel Size=72B, Type=Instruct2025.12 | 69 | 75.68 | 69.57 | 85.29 | 60.87 | 50 | |
| Qwen2.5-VL-InstructModel Size=32B, Type=Instruct2025.12 | 65 | 72.97 | 63.04 | 82.35 | 47.83 | 51.52 | |
| Claude-3.7-SonnetType=Proprietary2025.12 | 60 | 59.46 | 60.87 | 70.59 | 63.04 | 46.97 | |
| VeBrainParams=7B2026.06 | 58 | — | — | — | — | — | |
| Qwen2.5-VL-InstructModel Size=3B, Type=Instruct2025.12 | 56.67 | 63.51 | 52.17 | 66.18 | 47.83 | 48.48 | |
| Qwen2.5-VL-InstructModel Size=7B, Type=Instruct2025.12 | 56 | 50 | 78.26 | 70.59 | 50 | 36.36 | |
| InternVL3.5Params=8B2026.06 | 55.3 | — | — | — | — | — | |
| Mimo-EmbodiedParams=7B2026.06 | 54.6 | — | — | — | — | — | |
| Pelican-VLParams=7B2026.06 | 54.6 | — | — | — | — | — | |
| GPT-4oType=Proprietary2025.12 | 53.67 | 54.05 | 41.3 | 54.41 | 65.22 | 53.03 |