Spatial Reasoning on CV-Bench 3D
87.58CV-Bench 3D AccuracyQwen2.5-VL
Evaluation Results
| Method | Links | |
|---|---|---|
| Qwen2.5-VLSize=72B2026.07 | 87.58 | |
| GPT-5.2Size=–2026.07 | 86.92 | |
| BAGELSize=7B2026.07 | 83.41 | |
| GPT-4oSize=–2026.07 | 83.08 | |
| BRAIDSize=7B2026.07 | 82.17 | |
| SFTSize=7B2026.07 | 81.67 | |
| Qwen2.5-VLSize=7B2026.07 | 74.25 | |
| Janus-ProSize=7B2026.07 | 66.5 | |
| ChameleonSize=7B2026.07 | 46.58 |