Spatial Reasoning (Granular Metrics) on MMSI-Bench
49.5Average AccuracyMSSR
Evaluation Results
| Method | Links | ||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| MSSRModel Category=Agentic, Backbone=GPT-4o, Evaluation Protocol=zero-shot2025.10 | 49.5 | — | — | — | — | — | — | — | — | — | — | 50 | 50.6 | 47.1 | — | — | |
| Gemini-3-FlashWorkflow Mode=Forced Tool-Use (zero-shot), Type=Proprietary2026.06 | 49.5 | — | — | — | — | — | — | — | — | — | — | 50.3 | 50.4 | — | 51.5 | 43.4 | |
| S-AgentModel Category=Ours2026.06 | 46.4 | 46.2 | 43.6 | 37 | 43 | 43.5 | 63.9 | 57.8 | 40.9 | 46 | 48.7 | 44.4 | — | — | — | — | |
| Gemini-3-FlashWorkflow Mode=Direct Answer, Type=Proprietary2026.06 | 45.4 | — | — | — | — | — | — | — | — | — | — | 46 | 45.6 | — | 45.4 | 44 | |
| Gemini 3 ProModel Category=Proprietary Models2026.06 | 45.2 | 47.3 | 48.9 | 42 | 43 | 37.6 | 60.2 | 64.1 | 39.4 | 41.9 | 47.4 | 37.9 | — | — | — | — | |
| GPT-5Model Category=Proprietary2026.03 | 41.9 | 43 | 35.1 | 32.1 | 48.8 | 42.4 | 51.8 | 60.9 | 36.4 | 32.4 | 36.8 | 42 | — | — | — | — | |
| GPT-5.4Model Category=Proprietary Models2026.06 | 41.9 | 41.9 | 33 | 35.8 | 49.8 | 42.4 | 68.7 | 54.7 | 37.4 | 28.3 | 40.8 | 36.4 | — | — | — | — | |
| o3Model Category=Proprietary2026.03 | 41 | 45.2 | 39.4 | 37 | 44.2 | 47.1 | 62.6 | 54.7 | 28.8 | 31.1 | 32.9 | 34.9 | — | — | — | — | |
| o3Model Category=Proprietary LLM2025.10 | 41 | — | — | — | — | — | — | — | — | — | — | 34.9 | 45.8 | 36.4 | — | — | |
| GPT-4.5Model Category=Proprietary2026.03 | 40.3 | 34.4 | 29.8 | 39.5 | 51.2 | 47.1 | 55.4 | 39.1 | 33.3 | 41.9 | 40.8 | 36.4 | — | — | — | — | |
| LEOModel Category=Specialist2025.10 | 39.3 | — | — | — | — | — | — | — | — | — | — | 32.3 | 42.3 | 38.6 | — | — | |
| Astra Qwen3-VL-8B-InstructWorkflow Mode=Agentic Tool-Use, Type=Open-source2026.06 | 38.8 | — | — | — | — | — | — | — | — | — | — | 33.6 | 42.3 | — | 41 | 32.1 | |
| Seed 1.6Model Category=Open-weight General Models2026.06 | 38.5 | 36.6 | 36.2 | 32.1 | 32.6 | 42.4 | 46.9 | 48.4 | 33 | 31.1 | 42.1 | 40.4 | — | — | — | — | |
| SN-SI-1.1-Qwen3VL-8BModel Category=Open-weight Spatial Models2026.06 | 38.1 | 44.1 | 38.3 | 33.3 | 65.1 | 38.8 | 59 | 48.4 | 24.2 | 29.7 | 34.2 | 22.2 | — | — | — | — | |
| Gemini 2.5 ProModel Category=Proprietary Models2026.06 | 38 | 38.7 | 34 | 40.7 | 44.2 | 38.8 | 41 | 62.5 | 30.3 | 39.2 | 25 | 33.3 | — | — | — | — | |
| SN-U1-8B-MoTModel Category=Open-weight General Models2026.06 | 38 | 46.2 | 41.5 | 29.6 | 58.1 | 38.8 | 63.9 | 43.8 | 21.2 | 25.7 | 31.6 | 26.8 | — | — | — | — | |
| Grok 4Model Category=Proprietary Models2026.06 | 37.8 | 36.6 | 35.1 | 39.5 | 34.9 | 45.9 | 50.6 | 21.9 | 22.7 | 40.5 | 43.4 | 38.4 | — | — | — | — | |
| pySpatialModel Category=Baseline, Base Model=GPT-4o2026.03 | 37.3 | 51.6 | 28.7 | 27.2 | 20.9 | 41.2 | 38.6 | 46.9 | 39.4 | 46 | 38.2 | 36.4 | — | — | — | — | |
| Gemini 2.5 ProModel Category=Proprietary LLM2025.10 | 37 | — | — | — | — | — | — | — | — | — | — | 34.3 | 38.5 | 36.1 | — | — | |
| Gemini-2.5-ProModel Category=Proprietary2026.03 | 36.9 | 39.7 | 31.9 | 39.5 | 45.3 | 35.2 | 43.3 | 51.5 | 21.2 | 36.4 | 30.2 | 34.3 | — | — | — | — | |
| Gemini-2.5-ProWorkflow Mode=Direct Answer, Type=Proprietary2026.06 | 36.9 | — | — | — | — | — | — | — | — | — | — | 34.3 | 39 | — | 36.2 | 33.3 | |
| Qwen3.5-9BModel Category=Open-weight General Models2026.06 | 36.5 | 34.4 | 36.2 | 34.6 | 39.5 | 38.8 | 54.2 | 56.3 | 28.8 | 36.5 | 26.3 | 28.8 | — | — | — | — | |
| Gemini 2.5 FlashModel Category=Proprietary LLM2025.10 | 35 | — | — | — | — | — | — | — | — | — | — | 30.3 | 37.4 | 33.9 | — | — | |
| GLM-4.5VWorkflow Mode=Direct Answer, Type=Proprietary2026.06 | 33.8 | — | — | — | — | — | — | — | — | — | — | 30.3 | 35.6 | — | 36.9 | 29.3 | |
| SN-SI-1.1-Qwen2.5VL-7BModel Category=Open-weight Spatial Models2026.06 | 32.8 | 51.6 | 29.8 | 32.1 | 50 | 29.4 | 42.2 | 37.5 | 28.8 | 23 | 34.2 | 18.7 | — | — | — | — | |
| VST-7B-SFTModel Category=Open-weight Spatial Models2026.06 | 32.5 | 39.8 | 36.2 | 35.8 | 37.2 | 29.4 | 33.7 | 29.7 | 47 | 36.5 | 35.5 | 18.2 | — | — | — | — | |
| VLM-3RModel Category=3D-VLM2025.10 | 32 | — | — | — | — | — | — | — | — | — | — | 30.3 | 33 | 31.4 | — | — | |
| Qwen3-VL-4BModel Category=Open-source LLM2025.10 | 31.1 | — | — | — | — | — | — | — | — | — | — | 27.8 | 32.8 | 30.3 | — | — | |
| Qwen3-VL-8BModel Category=Open-source LLM2025.10 | 31.1 | — | — | — | — | — | — | — | — | — | — | 30.3 | 32.6 | 28.9 | — | — | |
| Qwen3-VL-8B-InstructModel Category=Open-weight General Models2026.06 | 31.1 | 28 | 37.2 | 32.1 | 31.4 | 35.3 | 38.5 | 37.5 | 15.2 | 27 | 28.9 | 29.8 | — | — | — | — | |
| Bagel-7B-MoTWorkflow Mode=Direct Answer, Type=Open-source2026.06 | 31 | — | — | — | — | — | — | — | — | — | — | 30.8 | 33.5 | — | 27.7 | 25.3 | |
| GPT-4.1Model Category=Proprietary2026.03 | 30.9 | 36.6 | 26.6 | 27.2 | 29.1 | 36.5 | 27.7 | 37.5 | 24.2 | 36.5 | 32.9 | 28.8 | — | — | — | — | |
| Qwen2.5-VL-72BModel Category=Open-source2026.03 | 30.7 | 25.8 | 34 | 34.6 | 23.3 | 34.1 | 36.1 | 45.3 | 27.3 | 27 | 30.3 | 27.3 | — | — | — | — | |
| Qwen2.5-VL-72BModel Category=Open-source LLM2025.10 | 30.7 | — | — | — | — | — | — | — | — | — | — | 27.3 | 31.2 | 32.1 | — | — | |
| Qwen3-VL-30B-InstructWorkflow Mode=Direct Answer, Type=Open-source2026.06 | 30.6 | — | — | — | — | — | — | — | — | — | — | 29.1 | 31.2 | — | 35.8 | 25.9 | |
| GPT-4oModel Category=Proprietary2026.03 | 30.3 | 34.4 | 24.5 | 23.5 | 19.8 | 37.6 | 27.7 | 32.8 | 31.8 | 35.1 | 36.8 | 30.8 | — | — | — | — | |
| GPT-4oModel Category=Baseline2026.03 | 30.3 | 34.4 | 24.5 | 23.5 | 19.8 | 37.6 | 27.7 | 32.8 | 31.8 | 35.1 | 36.8 | 30.8 | — | — | — | — | |
| GPT-4oModel Category=Proprietary LLM2025.10 | 30.3 | — | — | — | — | — | — | — | — | — | — | 30.8 | 28 | 34.3 | — | — | |
| GPT-4oWorkflow Mode=Direct Answer, Type=Proprietary2026.06 | 30.3 | — | — | — | — | — | — | — | — | — | — | 30.8 | 28 | — | 32.3 | 36 | |
| ViLaSRModel Category=Agentic, Evaluation Protocol=zero-shot2025.10 | 30.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Qwen3-VL-8B-InstructWorkflow Mode=Direct Answer, Type=Open-source2026.06 | 29.8 | — | — | — | — | — | — | — | — | — | — | 28.1 | 30.8 | — | 30.1 | 27.7 | |
| VILASRWorkflow Mode=Direct Answer, Type=Open-source2026.06 | 29.8 | — | — | — | — | — | — | — | — | — | — | 23.2 | 35.9 | — | 26 | 21 | |
| Bagel-7B-MoTWorkflow Mode=Forced Tool-Use (zero-shot), Type=Open-source2026.06 | 29.7 | — | — | — | — | — | — | — | — | — | — | 28.7 | 31.3 | — | 25.6 | 24.7 | |
| Qwen3-VL-30B-A3B-ThinkingModel Category=Open-weight General Models2026.06 | 29.4 | 23.7 | 31.9 | 35.8 | 31.4 | 36.5 | 22.9 | 40.6 | 19.7 | 18.9 | 27.6 | 31.3 | — | — | — | — | |
| InternVL3_5-8BModel Category=Open-weight General Models2026.06 | 29 | 29 | 26.6 | 29.6 | 24.4 | 31.8 | 25.3 | 29.7 | 25.8 | 14.9 | 34.2 | 36.4 | — | — | — | — | |
| VADARModel Category=Agentic, Evaluation Protocol=zero-shot2025.10 | 28.9 | — | — | — | — | — | — | — | — | — | — | 22.7 | 32.8 | 26.1 | — | — | |
| RoboBrain-2.0Workflow Mode=Direct Answer, Type=Open-source2026.06 | 28.9 | — | — | — | — | — | — | — | — | — | — | 28 | 28.9 | — | 28.8 | 22.5 | |
| Qwen3-VL-30B-InstructWorkflow Mode=Forced Tool-Use (zero-shot), Type=Open-source2026.06 | 28.9 | — | — | — | — | — | — | — | — | — | — | 28.1 | 31.5 | — | 28.7 | 21.6 | |
| InternVL2.5-8BModel Category=Open-source LLM2025.10 | 28.7 | — | — | — | — | — | — | — | — | — | — | 30.3 | 29.9 | 25.4 | — | — | |
| Qwen3-VL-8B-InstructWorkflow Mode=Forced Tool-Use (zero-shot), Type=Open-source2026.06 | 28.6 | — | — | — | — | — | — | — | — | — | — | 30.8 | 30.4 | — | 29.5 | 19.6 | |
| Qwen3-VL-8B-ThinkingModel Category=Open-weight General Models2026.06 | 28.6 | 31.2 | 26.6 | 32.1 | 29.1 | 32.9 | 30.1 | 50 | 16.7 | 17.6 | 23.7 | 27.3 | — | — | — | — | |
| InternVL3-78BModel Category=Open-source2026.03 | 28.5 | 34.4 | 23.4 | 32.1 | 12.8 | 37.6 | 26.5 | 37.5 | 19.7 | 28.4 | 31.6 | 29.3 | — | — | — | — | |
| InternVL2.5-78BModel Category=Open-source2026.03 | 28.5 | 23.7 | 22.3 | 39.5 | 29.1 | 31.8 | 42.2 | 35.9 | 19.7 | 17.6 | 26.3 | 27.3 | — | — | — | — | |
| LLaVA-OneVision-72BModel Category=Open-source2026.03 | 28.4 | 43 | 31.9 | 33.3 | 30.2 | 37.6 | 38.6 | 28.1 | 19.7 | 13.5 | 32.9 | 15.7 | — | — | — | — | |
| LLaVA-3DModel Category=3D-VLM2025.10 | 28 | — | — | — | — | — | — | — | — | — | — | 19.7 | 29.5 | 31 | — | — | |
| Video-R1Workflow Mode=Direct Answer, Type=Open-source2026.06 | 27.8 | — | — | — | — | — | — | — | — | — | — | 26.8 | 30.5 | — | 25.4 | 22 | |
| VLaserWorkflow Mode=Direct Answer, Type=Open-source2026.06 | 27.3 | — | — | — | — | — | — | — | — | — | — | 18.9 | 29.8 | — | 26.9 | 26 | |
| SpaceRWorkflow Mode=Direct Answer, Type=Open-source2026.06 | 26.9 | — | — | — | — | — | — | — | — | — | — | 22.5 | 29.1 | — | 29.4 | 21.9 | |
| InternVL3-14BModel Category=Open-source LLM2025.10 | 26.8 | — | — | — | — | — | — | — | — | — | — | 29.3 | 25.5 | 27.5 | — | — | |
| Qwen2.5-VL-3BModel Category=Open-source LLM2025.10 | 26.5 | — | — | — | — | — | — | — | — | — | — | 23.2 | 29.5 | 23.2 | — | — | |
| Spatial-MLLMWorkflow Mode=Direct Answer, Type=Open-source2026.06 | 26.1 | — | — | — | — | — | — | — | — | — | — | 26.3 | 28.5 | — | 25.4 | 18 | |
| Qwen2.5-VL-7BModel Category=Open-source LLM2025.10 | 25.9 | — | — | — | — | — | — | — | — | — | — | 25.8 | 25.9 | 26.1 | — | — | |
| SpatialLadderWorkflow Mode=Direct Answer, Type=Open-source2026.06 | 25.4 | — | — | — | — | — | — | — | — | — | — | 21.2 | 30.3 | — | 23.3 | 16 | |
| Video 3D LLMModel Category=3D-VLM2025.10 | 25.3 | — | — | — | — | — | — | — | — | — | — | 25.8 | 25.5 | 23.9 | — | — | |
| SpatialLLMWorkflow Mode=Direct Answer, Type=Open-source2026.06 | 25.3 | — | — | — | — | — | — | — | — | — | — | 30.8 | 24.5 | — | 23.1 | 22.7 | |
| Llama-3.2-11B-VisionModel Category=Open-source LLM2025.10 | 24.5 | — | — | — | — | — | — | — | — | — | — | 19.2 | 26.9 | 23.9 | — | — | |
| LLaVA-OneVision-7BModel Category=Open-source LLM2025.10 | 24.5 | — | — | — | — | — | — | — | — | — | — | 11.6 | 28 | 27.1 | — | — | |
| Skill-3DModel=GPT-4o2026.06 | — | — | — | — | — | — | — | — | — | — | — | — | 43.2 | — | — | — | |
| Skill-3DModel=GPT-5.42026.06 | — | — | — | — | — | — | — | — | — | — | — | — | 60.4 | — | — | — | |
| Skill-3DModel=Gemini-2.5-Pro2026.06 | — | — | — | — | — | — | — | — | — | — | — | — | 56.7 | — | — | — | |
| Skill-3DModel=Gemini-3-Flash2026.06 | — | — | — | — | — | — | — | — | — | — | — | — | 54.8 | — | — | — | |
| Think3DModel=GPT-4o2026.06 | — | — | — | — | — | — | — | — | — | — | — | — | 38.4 | — | — | — | |
| Think3DModel=GPT-5.42026.06 | — | — | — | — | — | — | — | — | — | — | — | — | 53.4 | — | — | — | |
| Think3DModel=Gemini-2.5-Pro2026.06 | — | — | — | — | — | — | — | — | — | — | — | — | 51 | — | — | — | |
| Think3DModel=Gemini-3-Flash2026.06 | — | — | — | — | — | — | — | — | — | — | — | — | 49.2 | — | — | — | |
| w/ ToolsModel=GPT-4o2026.06 | — | — | — | — | — | — | — | — | — | — | — | — | 34.6 | — | — | — | |
| w/ ToolsModel=GPT-5.42026.06 | — | — | — | — | — | — | — | — | — | — | — | — | 48.1 | — | — | — | |
| w/ ToolsModel=Gemini-2.5-Pro2026.06 | — | — | — | — | — | — | — | — | — | — | — | — | 44.3 | — | — | — | |
| w/ ToolsModel=Gemini-3-Flash2026.06 | — | — | — | — | — | — | — | — | — | — | — | — | 36.8 | — | — | — | |
| w/o ToolsModel=GPT-4o2026.06 | — | — | — | — | — | — | — | — | — | — | — | — | 31.1 | — | — | — | |
| w/o ToolsModel=GPT-5.42026.06 | — | — | — | — | — | — | — | — | — | — | — | — | 42.7 | — | — | — | |
| w/o ToolsModel=Gemini-2.5-Pro2026.06 | — | — | — | — | — | — | — | — | — | — | — | — | 36.9 | — | — | — | |
| w/o ToolsModel=Gemini-3-Flash2026.06 | — | — | — | — | — | — | — | — | — | — | — | — | 32.7 | — | — | — |