Spatial Reasoning on SPAR-Bench
67.3Overall ScoreHuman Level
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| Human LevelModel Category=Baseline2026.05 | 67.3 | — | — | — | |
| SpatialImaginer2026.04 | 62.3 | — | — | — | |
| Cosmos3 Nano2026.07 | 54.8 | — | — | — | |
| GPT-52026.04 | 54.72 | — | — | — | |
| Ovis2.5Params=7B2026.04 | 51.44 | — | — | — | |
| SpatialForgeQwen3-VL-2BModel Category=Ours2026.05 | 50.6 | — | — | — | |
| RynnBrain 8B2026.07 | 49.8 | — | — | — | |
| Gemini 2.5 Pro2026.04 | 49.42 | — | — | — | |
| Gemini 2.5 Pro2026.07 | 46.2 | — | — | — | |
| GPT 5.42026.07 | 46.1 | — | — | — | |
| ACE-Brain 0-8B2026.07 | 44.2 | — | — | — | |
| GPT-4.12026.04 | 42.77 | — | — | — | |
| Qwen3-VL-2BModel Category=Open-Source General Models2026.05 | 42.6 | — | — | — | |
| RoboBrain 2.5-8B2026.07 | 42.5 | — | — | — | |
| RoboBrain 2.0-7B2026.07 | 42.5 | — | — | — | |
| VLM-3R-7B2026.04 | 42.4 | — | — | — | |
| SpaceVistaParams=7B, Data Size=1M2026.04 | 41.6 | — | — | — | |
| Vlaser 8B2026.07 | 41.2 | — | — | — | |
| SpaceMantis-8BModel Category=Open-sourced Specialized Models2026.05 | 41 | — | — | — | |
| Embodied R1.52026.07 | 40.3 | — | — | — | |
| ACE-Brain 0.5-8B2026.07 | 39.7 | — | — | — | |
| Qwen2.5-VL-7BModel Category=Open-Source General Models2026.05 | 39.2 | — | — | — | |
| EgoMindParams=7B, Data Size=25k2026.04 | 39.03 | — | — | — | |
| Pelican VL-7B2026.07 | 38.7 | — | — | — | |
| SpaceRParams=7B, Data Size=151k2026.04 | 38.26 | — | — | — | |
| InternVL3.5Params=7B2026.04 | 38.05 | — | — | — | |
| Cambrian-S-7B2026.04 | 37.9 | — | — | — | |
| SpaceR-7BModel Category=Open-sourced Specialized Models2026.05 | 37.6 | — | — | — | |
| ViLaSR-7B2026.04 | 37.4 | — | — | — | |
| SpaceR-7BBase Model=Qwen2.5-VL-7B2025.11 | 37.3 | 37.91 | 36.82 | — | |
| SpaceQwen2.5-VL-3B-InstructModel Category=Open-sourced Specialized Models2026.05 | 36.9 | — | — | — | |
| V2LO-7BReinforcement Learning=true2025.11 | 36.68 | 36.63 | 36.72 | — | |
| GPT-4oModel Category=Proprietary Models2026.05 | 36.4 | — | — | — | |
| GPT-4oModel Type=Closed-source2025.11 | 36.39 | 38.1 | 35.3 | — | |
| InternVL2.5-8BModel Type=Open-source2025.11 | 36.28 | 35.96 | 36.46 | — | |
| MiniCPM-V-4.5Params=7B2026.04 | 35.98 | — | — | — | |
| Spatial-MLLM-4B2026.04 | 35.3 | — | — | — | |
| Spatial-MLLMParams=4B, Data Size=120K2026.04 | 35.1 | — | — | — | |
| ViLaSRParams=7B, Data Size=80k2026.04 | 34.98 | — | — | — | |
| V2LO-7BReinforcement Learning=false2025.11 | 34.66 | 33.51 | 35.41 | — | |
| SpatialLadder-3BModel Category=Open-sourced Specialized Models2026.05 | 34.4 | — | — | — | |
| SpaceR-7B2026.04 | 34.2 | — | — | — | |
| Qwen2.5-VLParams=7B2026.04 | 33.19 | — | — | — | |
| SpatialLadder-3B2026.04 | 32.9 | — | — | — | |
| Qwen2.5VL-7BBase Model=Qwen2.5-VL-7B2025.11 | 32.73 | 32.57 | 28.98 | — | |
| RandomModel Category=Baseline2026.05 | 32.7 | — | — | — | |
| VeBrain 7B2026.07 | 32.5 | — | — | — | |
| Spatial-MLLM-4BModel Category=Open-sourced Specialized Models2026.05 | 31.5 | — | — | — | |
| MiMo-VL-RLParams=7B2026.04 | 31.44 | — | — | — | |
| LLaVA-OneVision-7BModel Type=Open-source2025.11 | 31.2 | 33.13 | 29.92 | — | |
| LLaVA-OneVision-7BModel Category=Open-Source General Models2026.05 | 31.2 | — | — | — | |
| Claude Sonnet 4.62026.07 | 31.2 | — | — | — | |
| LLaVA-OneVisionParams=7B2026.04 | 30.6 | — | — | — | |
| Gemini 2.0 flashModel Type=Closed-source2025.11 | 30.08 | — | — | — | |
| GPT4SceneParams=7B2026.04 | 26.53 | — | — | — | |
| Qwen2.5-VL-3BModel Category=Open-Source General Models2026.05 | 24.6 | — | — | — | |
| Claude-3.7-SonnetModel Category=Proprietary Models2026.05 | 21.8 | — | — | — | |
| MindCube-3B2026.04 | 20.8 | — | — | — | |
| MiMo Embodied-7B2026.07 | 11.1 | — | — | — | |
| BAGEL-7B-MoTCategory=Open-source General Models2026.05 | — | — | — | 39.1 | |
| Cambrian-S-7BScale=7B2026.05 | — | — | — | 37.9 | |
| Cambrian-S-7BCategory=Open-source Spatial Intelligence Models2026.05 | — | — | — | 37.9 | |
| Gemini 2.5 ProCategory=Proprietary Models2026.05 | — | — | — | 49.8 | |
| Gemini 3 ProCategory=Proprietary Models2026.05 | — | — | — | 48.7 | |
| GeoThinkerCategory=Open-source Spatial Intelligence Models2026.05 | — | — | — | 68.2 | |
| GeoWeaver-5BScale=5B2026.05 | — | — | — | 69.2 | |
| GPT-3.2Category=Proprietary Models2026.05 | — | — | — | 55.1 | |
| GPT-4o2026.05 | — | — | — | 38.1 | |
| GR3D-8BParameters=8B, Backbone=Qwen-2-7B, Input Resolution=4482026.05 | — | — | — | 43.7 | |
| HumanCategory=Reference2026.05 | — | — | — | 67.3 | |
| InternVL2.5-8BParameters=8B2026.05 | — | — | — | 36.3 | |
| InternVL3.5-8BCategory=Open-source General Models2026.05 | — | — | — | 38.2 | |
| MindCube-3BScale=3B2026.05 | — | — | — | 20.8 | |
| NVILA-8BParameters=8B2026.05 | — | — | — | 34.1 | |
| ProSRDescription=QWEN3-VL-8B-THINKING + Ours2026.05 | — | — | — | 64.3 | |
| Qwen2.5-VL-7BParameters=7B2026.05 | — | — | — | 33.1 | |
| Qwen2.5-VL-7B-InstructCategory=Open-source General Models2026.05 | — | — | — | 33.8 | |
| Qwen3-VL-8BParameters=8B2026.05 | — | — | — | 39.8 | |
| Qwen3-VL-8B-InstructCategory=Open-source General Models2026.05 | — | — | — | 39.6 | |
| Qwen3-VL-8B-ThinkingCategory=Open-source General Models2026.05 | — | — | — | 33 | |
| QWEN3-VL-8B-THINKING + GRPOMethod Variant=Vanilla GRPO2026.05 | — | — | — | 58.4 | |
| QWEN3-VL-8B-THINKING + SFTMethod Variant=SFT Only2026.05 | — | — | — | 54.3 | |
| Random ChoiceCategory=Reference2026.05 | — | — | — | 32.7 | |
| Seed 1.6Category=Proprietary Models2026.05 | — | — | — | 50.1 | |
| SenseNova-SI-1.1-Qwen3-VL-8BCategory=Open-source Spatial Intelligence Models2026.05 | — | — | — | 40.8 | |
| SpaceR-7BScale=7B2026.05 | — | — | — | 34.2 | |
| SpaceR-SFT-7BCategory=Open-source Spatial Intelligence Models2026.05 | — | — | — | 34.2 | |
| Spatial-MLLM-4BScale=4B2026.05 | — | — | — | 35.3 | |
| SpatialLadder-3BScale=3B2026.05 | — | — | — | 32.9 | |
| SpatialLadder-3BCategory=Open-source Spatial Intelligence Models2026.05 | — | — | — | 32.9 | |
| SR-3D-8BParameters=8B2026.05 | — | — | — | 32.1 | |
| ViLaSR-7BScale=7B2026.05 | — | — | — | 37.4 | |
| VLM-3R-7BScale=7B2026.05 | — | — | — | 42.4 | |
| VST-7B-SFTCategory=Open-source Spatial Intelligence Models2026.05 | — | — | — | 46.6 |