Spatial Understanding on CV-Bench
88.82AccuracyMiMo-Embodied
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| MiMo-EmbodiedParams=7B2025.11 | 88.82 | — | — | — | |
| GEM-8BParameters=8B2026.05 | 86.6 | — | — | — | |
| Seed1.82026.05 | 86.5 | — | — | — | |
| RoboBrain-2.0Params=7B2025.11 | 85.75 | — | — | — | |
| Qwen3-VL-8B-SFTParameters=8B, Training=SFT2026.05 | 85.6 | — | — | — | |
| Qwen3-VL-8BParameters=8B2026.05 | 85.1 | — | — | — | |
| Gemini2.5-ProParams=–2025.11 | 84.59 | — | — | — | |
| VST-7BParameters=7B2026.05 | 83.5 | — | — | — | |
| SenseNova-SI-8BParameters=8B2026.05 | 83.2 | — | — | — | |
| Gemini-3-Pro2026.05 | 82.5 | — | — | — | |
| MiMo-VLParams=7B2025.11 | 81.69 | — | — | — | |
| InternVL3.5-8BParameters=8B2026.05 | 81.5 | — | — | — | |
| InternVL3.5Params=8B2025.11 | 81.46 | — | — | — | |
| GEM-2BParameters=2B2026.05 | 81.4 | — | — | — | |
| Qwen3-VL-2B-SFTParameters=2B, Training=SFT2026.05 | 80.7 | — | — | — | |
| Qwen3-VL-2BParameters=2B2026.05 | 80 | — | — | — | |
| VeBrainParams=8B2025.11 | 79.68 | — | — | — | |
| Qwen-VL-MaxParams=–2025.11 | 79.42 | — | — | — | |
| GPT-4oParams=–2025.11 | 78.63 | — | — | — | |
| GPT-4o2026.05 | 78.6 | — | — | — | |
| Claude-Sonnet-4Params=–2025.11 | 78.12 | — | — | — | |
| CambrainS-7BParameters=7B2026.05 | 76.9 | — | — | — | |
| RoboBrain-1.0Params=7B2025.11 | 76.22 | — | — | — | |
| Qwen2.5-VLParams=7B2025.11 | 75.4 | — | — | — | |
| SpaceR-7BParameters=7B2026.05 | 74.8 | — | — | — | |
| VLM-3R2026.05 | 71.8 | — | — | — | |
| Cosmos-Reason1Params=7B2025.11 | 68.57 | — | — | — | |
| MagmaParams=8B2025.11 | 65.88 | — | — | — | |
| LLava-OneVision-7BParameters=7B2026.05 | 61.9 | — | — | — | |
| Cambrian-1-8BModel Scale=8B2025.11 | — | 72.3 | 72 | 72.2 | |
| Claude-3.5-Sonnet-0620Model Type=Proprietary2025.11 | — | 60.2 | 71.5 | 65.9 | |
| Claude-4-Sonnet-0514Model Type=Proprietary2025.11 | — | 73.3 | 84.2 | 78.7 | |
| GPT-4o-0513Model Type=Proprietary2025.11 | — | 75.8 | 83 | 79.4 | |
| GPT-5-0807Model Type=Proprietary2025.11 | — | 81.4 | 90.3 | 85.8 | |
| LLaVA-NeXT-8BModel Scale=8B2025.11 | — | 62.2 | 65.3 | 63.8 | |
| Qwen2.5-VL-3BModel Scale=3B2025.11 | — | 59.9 | 60.2 | 60 | |
| Qwen2.5-VL-3B + SFTModel Scale=3B, Training Protocol=SFT2025.11 | — | 53.9 | 68.4 | 61.1 | |
| Qwen2.5-VL-3B + Vanilla GRPOModel Scale=3B, Training Protocol=Vanilla GRPO2025.11 | — | 70.6 | 66.6 | 68.6 | |
| Qwen2.5-VL-7BModel Scale=7B2025.11 | — | 69.1 | 68 | 68.6 | |
| Qwen2.5-VL-7B + SFTModel Scale=7B, Training Protocol=SFT2025.11 | — | 56.1 | 71.3 | 63.7 | |
| Qwen2.5-VL-7B + Vanilla GRPOModel Scale=7B, Training Protocol=Vanilla GRPO2025.11 | — | 68.9 | 76.5 | 72.7 | |
| Qwen3-VL-30BModel Scale=30B2025.11 | — | 79 | 89.6 | 84.3 | |
| RoboPoint-13BModel Scale=13B2025.11 | — | — | 61.2 | — | |
| SATORI-R12025.11 | — | 54.6 | 69.4 | 62 | |
| SpaceLLaVA-13BModel Scale=13B2025.11 | — | — | 68.5 | — | |
| SpaceOm2025.11 | — | 72.1 | 69.3 | 70.7 | |
| SpaceThinker2025.11 | — | 65.1 | 65.9 | 65.5 | |
| Spatial-RGPT-7B w/ depthModel Scale=7B, Auxiliary Inputs=depth2025.11 | — | — | 60.7 | — | |
| SpatialBot-3BModel Scale=3B2025.11 | — | — | 69.1 | — | |
| SpatialThinker-30B (Ours)Model Scale=30B2025.11 | — | 80.3 | 93.6 | 87 | |
| SpatialThinker-3B (Ours)Model Scale=3B2025.11 | — | 71 | 76.3 | 73.6 | |
| SpatialThinker-7B (Ours)Model Scale=7B2025.11 | — | 77.7 | 78.7 | 78.2 | |
| VLAA-Thinker-Qwen2.5-VL-7BModel Scale=7B2025.11 | — | 60.8 | 60.3 | 60.6 |