Spatial Reasoning (Multi-Image) on MMSI-Bench
97.2AccuracyHuman
Evaluation results
| Method | Links | |
|---|---|---|
| Human2025.10 | 97.2 | |
| Gemini-3-Pro-PreviewSize=-2026.03 | 45.2 | |
| GPT-5Size=-2026.03 | 41.8 | |
| GPT-52025.10 | 40.7 | |
| Gemini-2.5-ProSize=-2026.03 | 38 | |
| Gemini-2.5-pro2025.10 | 36.9 | |
| InternVL3.5-38BParameters=38B2025.10 | 36.9 | |
| ViewFusionSize=4B, Training Protocol=SFT + RL2026.03 | 35.4 | |
| VST-7B-RLSize=7B, Training Protocol=RL2026.03 | 34.8 | |
| ViewFusionSize=4B, Training Protocol=SFT2026.03 | 32.4 | |
| SpaceVista-7BParameters=7B, RL=true2025.10 | 32.3 | |
| VST-3B-RLSize=3B, Training Protocol=RL2026.03 | 32 | |
| Qwen2.5-VL-7BParameters=7B2025.10 | 31.7 | |
| ViLAVT-7BThinking Level=Chatting with Images2026.02 | 31.3 | |
| Qwen3-VL-8B-InstructSize=8B, Training Protocol=Instruct2026.03 | 31.1 | |
| InternVL3.5-8BParameters=8B2025.10 | 30.9 | |
| Qwen2.5-VL-72BParameters=72B2025.10 | 30.7 | |
| VILASR-7BThinking Level=Thinking with Images2026.02 | 30.2 | |
| ViLaSR-7BSize=7B2026.03 | 30.2 | |
| VILASR-7BParameters=7B2025.10 | 30.2 | |
| Qwen3-VL-4B-InstructSize=4B, Training Protocol=Instruct2026.03 | 30.1 | |
| SpaceVista-7BParameters=7B2025.10 | 29.1 | |
| Qwen3-VL-2B-InstructSize=2B, Training Protocol=Instruct2026.03 | 28.9 | |
| SpaceR-7BThinking Level=Thinking about Images2026.02 | 28.8 | |
| Qwen2.5-VL-3B-InstructSize=3B, Training Protocol=Instruct2026.03 | 28.6 | |
| InternVL3-8BSize=8B2026.03 | 28 | |
| Spatial-MLLM-4BThinking Level=Thinking about Images2026.02 | 27.7 | |
| SpatialLadder-3BSize=3B2026.03 | 27.4 | |
| SpaceR-7BSize=7B2026.03 | 27.4 | |
| Qwen2.5-VL-7BFine-tuned on=SpaceVista-1M2025.10 | 27.3 | |
| Qwen2.5-VL-7BThinking Level=Thinking about Images2026.02 | 27.1 | |
| SpatialMLLM-4BParameters=4B2025.10 | 27 | |
| Qwen2.5-VL-7BThinking Level=Non-Thinking2026.02 | 26.9 | |
| Qwen2.5-VL-7B-InstructSize=7B, Training Protocol=Instruct2026.03 | 26.8 | |
| LLaVA-NeXT-Video-7BParameters=7B2025.10 | 26.8 | |
| InternVL3-2BSize=2B2026.03 | 26.5 | |
| Spatial-MLLM-4BSize=4B2026.03 | 26.1 | |
| SpaceR-7BParameters=7B2025.10 | 26.1 | |
| Cambrian-S-7BSize=7B2026.03 | 25.8 | |
| InternVL3-8BThinking Level=Non-Thinking2026.02 | 25.7 | |
| Cambrian-S-3BSize=3B2026.03 | 25.2 | |
| RandomChoiceSize=-2026.03 | 25 | |
| LLaVA-OneVision-7BThinking Level=Non-Thinking2026.02 | 24.5 | |
| LLAVA-Onevision-7BParameters=7B2025.10 | 24.5 |