Video-based spatial intelligence on MMSI-Video-Bench
38Average ScoreGemini 3 Pro
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| Gemini 3 ProModel Category=Proprietary Models2026.06 | 38 | — | — | — | — | |
| GPT-5Model Category=Proprietary Models2026.06 | 36.8 | — | — | — | — | |
| Gemini 2.5 FlashModel Category=Proprietary Models2026.06 | 35.4 | — | — | — | — | |
| ViSRABackbone=InternVL3-8B2026.05 | 34.5 | 29.9 | 24.6 | 47.4 | 35 | |
| ViSRABackbone=Qwen2.5-VL-7B2026.05 | 33.9 | 27.3 | 27.5 | 42.3 | 37.5 | |
| InternVideo3Model Category=Open-Weight Models2026.06 | 30.7 | — | — | — | — | |
| ViSRABackbone=LLaVA-OneVision-7B2026.05 | 29.2 | 22.1 | 29 | 23.1 | 42.1 | |
| InternVL3.5-8BModel Category=Open-Weight Models, Parameter count=8B2026.06 | 29.2 | — | — | — | — | |
| LLaVA-OneVision-7BType=Base Model2026.05 | 28.6 | 22.1 | 33.3 | 21.8 | 37.6 | |
| VideoChat-Flash-7BModel Category=Open-Weight Models, Parameter count=7B2026.06 | 28.2 | — | — | — | — | |
| Qwen3-VL-8BModel Category=Open-Weight Models, Parameter count=8B2026.06 | 27.6 | — | — | — | — | |
| InternVideo2.5-7BModel Category=Open-Weight Models, Parameter count=7B2026.06 | 27.3 | — | — | — | — | |
| Keye-VL-1.5-8BModel Category=Open-Weight Models, Parameter count=1.5-8B2026.06 | 26.5 | — | — | — | — | |
| InternVL3-8BType=Base Model2026.05 | 25 | 26 | 27.5 | 20.5 | 26.3 | |
| Qwen2.5-VL-7BType=Base Model2026.05 | 23.7 | 23.4 | 26.1 | 16.6 | 28.8 | |
| MiniCPM-V-4.5-8BModel Category=Open-Weight Models, Parameter count=4.5-8B2026.06 | 23.4 | — | — | — | — |