Video spatial & 4D reasoning on VSTI-Bench 2026
70.3AccuracySpatialClaw
Evaluation Results
| Method | Links | |
|---|---|---|
| SpatialClawModel=Qwen3.6-27B2026.06 | 70.3 | |
| SpatialClawModel=Gemma4-31B2026.06 | 67.6 | |
| SpatialClawModel=Qwen3.5-397B-A17B2026.06 | 67.5 | |
| SpatialClawModel=Qwen3.6-35B-A3B2026.06 | 64.2 | |
| SpatialClawModel=Qwen3.5-122B-A10B2026.06 | 63.3 | |
| SpatialClawModel=Gemma4-26B-A4B2026.06 | 61.7 | |
| No-toolModel=Qwen3.6-27B2026.06 | 60.5 | |
| No-toolModel=Qwen3.5-397B-A17B2026.06 | 60.4 | |
| No-toolModel=Qwen3.6-35B-A3B2026.06 | 58.8 | |
| No-toolModel=Qwen3.5-122B-A10B2026.06 | 54.7 | |
| No-toolModel=Gemma4-31B2026.06 | 54.7 | |
| No-toolModel=Gemma4-26B-A4B2026.06 | 53.1 |