Multi-image Spatial Reasoning on SPAR-Bench-MV (test)
43.7Score (Low Difficulty)GPT-5.2
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| GPT-5.2Category=Proprietary2026.02 | 43.7 | 41.5 | 66.4 | 52.6 | |
| HATCHBase Model=Qwen2.5-VL-3B2026.02 | 41.3 | 47.4 | 67.1 | 53.6 | |
| Gemini-3-ProCategory=Proprietary2026.02 | 40.6 | 21.7 | 58 | 43.1 | |
| GPT-4.1Category=Proprietary2026.02 | 38.2 | 37 | 43.1 | 39.9 | |
| InternVL-2.5-8BCategory=Open-Weight2026.02 | 31.8 | 30.8 | 48.2 | 38.3 | |
| SpaceR-7BBase Model=Qwen2.5-VL-7B2026.02 | 31.6 | 32.3 | 49.2 | 39.1 | |
| InternVL-2.5-4BCategory=Open-Weight2026.02 | 30.3 | 30.1 | 38.5 | 33.6 | |
| Video-R1Base Model=Qwen2.5-VL-7B2026.02 | 29.4 | 30 | 46.2 | 36.5 | |
| Qwen2.5-VL-72BCategory=Open-Weight2026.02 | 26.9 | 32.7 | 43.6 | 35.4 | |
| SpatialLadder-3BBase Model=Qwen2.5-VL-3B2026.02 | 26.2 | 33.5 | 44.5 | 35.8 | |
| Qwen2.5-VL-32BCategory=Open-Weight2026.02 | 25.6 | 29 | 45.4 | 34.7 | |
| LLaVA-OneVision-4BCategory=Open-Weight2026.02 | 23 | 28 | 40.1 | 31.5 | |
| Spatial-MLLM-4BBase Model=Qwen2.5-VL-3B2026.02 | 22.3 | 32.4 | 35.2 | 30.4 | |
| Qwen2.5-VL-7BBase Model=Qwen2.5-VL-7B2026.02 | 16.9 | 29.4 | 40.7 | 30.1 | |
| Qwen2.5-VL-3BBase Model=Qwen2.5-VL-3B2026.02 | 13.4 | 26.2 | 32.6 | 24.9 |