Spatial Reasoning on Spatial reasoning benchmark
72.9Overall AccuracyCAMCRAFT
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| CAMCRAFTUnderstanding Model=GPT-5, Image Editor=CAMCRAFT2026.05 | 72.9 | 65.3 | 82.6 | 91.1 | 66.2 | |
| GPTIMAGE1Understanding Model=GPT-5, Image Editor=GPTIMAGE12026.05 | 71.7 | 65.3 | 80.4 | 91.1 | 64.1 | |
| NANOBANANAPROUnderstanding Model=GPT-5, Image Editor=NANOBANANAPRO2026.05 | 70.9 | 60.9 | 83 | 91.1 | 63.6 | |
| w/o GenUnderstanding Model=GPT-5, Image Editor=None2026.05 | 70.4 | 63.6 | 80.9 | 93.3 | 60.5 | |
| QWENIMAGEEDITUnderstanding Model=GPT-5, Image Editor=QWENIMAGEEDIT2026.05 | 68.6 | 61.8 | 77.8 | 86.7 | 61.5 | |
| GeoWeaver-5BScale=5B2026.05 | 64.8 | — | — | — | — | |
| VLM-3R-7BScale=7B2026.05 | 52.6 | — | — | — | — | |
| Cambrian-S-7BScale=7B2026.05 | 49.6 | — | — | — | — | |
| ViLaSR-7BScale=7B2026.05 | 47.2 | — | — | — | — | |
| SpaceR-7BScale=7B2026.05 | 44.8 | — | — | — | — | |
| SpatialLadder-3BScale=3B2026.05 | 43.6 | — | — | — | — | |
| Spatial-MLLM-4BScale=4B2026.05 | 40.5 | — | — | — | — | |
| MindCube-3BScale=3B2026.05 | 22.8 | — | — | — | — |