Spatial Reasoning on ViewSpatial-Bench
51.8Overall ScoreMSSR
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| MSSRModel Category=Agentic, Backbone=GPT-4o, Evaluation Protocol=zero-shot2025.10 | 51.8 | 51 | 54.4 | |
| o3Model Category=Proprietary LLM2025.10 | 51.1 | 51.3 | 51 | |
| LEOModel Category=Specialist2025.10 | 43.7 | 41.5 | 45.8 | |
| InternVL2.5-8BModel Category=Open-source LLM2025.10 | 43.2 | 46.5 | 40.2 | |
| Qwen2.5-VL-72BModel Category=Open-source LLM2025.10 | 43.1 | 47.6 | 38.9 | |
| Gemini 2.5 ProModel Category=Proprietary LLM2025.10 | 43 | 44.3 | 41.6 | |
| Qwen3-VL-8BModel Category=Open-source LLM2025.10 | 41.2 | 45.1 | 37.5 | |
| CaMo-3BParameters=3B2026.05 | 41 | 41.5 | 40.4 | |
| InternVL3-14BModel Category=Open-source LLM2025.10 | 40.3 | 47.1 | 33.9 | |
| V2LO-7BReinforcement Learning=true2025.11 | 40.18 | 41.66 | 38.78 | |
| VLM-3RModel Category=3D-VLM2025.10 | 39.2 | 43.8 | 34.9 | |
| Qwen3-VL-4BModel Category=Open-source LLM2025.10 | 39 | 42.3 | 35.9 | |
| Gemini 2.5 FlashModel Category=Proprietary LLM2025.10 | 38.4 | 41.6 | 36.9 | |
| SpaceR-7BBase Model=Qwen2.5-VL-7B2025.11 | 38.38 | 41.1 | 35.81 | |
| ViLaSRModel Category=Agentic, Evaluation Protocol=zero-shot2025.10 | 38.2 | 42.4 | 34.4 | |
| InternVL2.5-8BModel Type=Open-source2025.11 | 38.13 | 36.8 | 39.38 | |
| V2LO-7BReinforcement Learning=false2025.11 | 38.13 | 39.13 | 36.83 | |
| InternVL-2.5-4BParameters=4B2026.05 | 37.9 | 40.8 | 35.1 | |
| Qwen2.5-VL-7BParameters=7B2026.05 | 37.9 | 41.8 | 34.3 | |
| Qwen2.5VL-7BBase Model=Qwen2.5-VL-7B2025.11 | 37.11 | 39.54 | 34.83 | |
| Qwen2.5-VL-7BModel Category=Open-source LLM2025.10 | 36.9 | 40.6 | 33.4 | |
| Qwen2.5-VL-3BModel Category=Open-source LLM2025.10 | 35.8 | 39.8 | 32.1 | |
| Qwen2.5-VL-3BParameters=3B2026.05 | 35.6 | 39.5 | 32 | |
| GPT-4oModel Category=Proprietary LLM2025.10 | 35 | 33.6 | 36.3 | |
| GPT-4oModel Type=Closed-source2025.11 | 34.95 | — | — | |
| LLaVA-3DModel Category=3D-VLM2025.10 | 33.8 | 39.6 | 28.3 | |
| VADARModel Category=Agentic, Evaluation Protocol=zero-shot2025.10 | 33.7 | 34.2 | 33.2 | |
| Kimi-VL-A3BParameters=3B2026.05 | 33.6 | 25.1 | 41.5 | |
| GPT-4oAccess=Proprietary2026.05 | 32.6 | 33.7 | 31.5 | |
| Gemini 2.0 flashModel Type=Closed-source2025.11 | 32.56 | — | — | |
| Video 3D LLMModel Category=3D-VLM2025.10 | 30.3 | 24.4 | 35.8 | |
| Llama-3.2-11B-VisionModel Category=Open-source LLM2025.10 | 28.8 | 23.7 | 33.6 | |
| LLaVA-OneVision-7BModel Category=Open-source LLM2025.10 | 27.5 | 28.5 | 26.6 | |
| LLaVA-OneVision-7BParameters=7B2026.05 | 27.5 | 28.5 | 26.5 | |
| LLaVA-OneVision-7BModel Type=Open-source2025.11 | 27.49 | 28.49 | 26.54 |