Social Understanding on OmniSocialBench
58.13Social Norm AccuracyCogniRoute
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| CogniRouteModel Category=CogniRoute model2026.06 | 58.13 | 66.44 | 60.18 | 55.71 | 59.38 | |
| Gemini 3.1 ProModel Category=VLMs that take omni inputs2026.06 | 33.33 | 36.84 | 60 | 43.9 | 44.05 | |
| Qwen 3.5 OmniModel Category=VLMs that take omni inputs2026.06 | 31.25 | 30.87 | 36.02 | 28.57 | 31.5 | |
| MiMo V2 OmniModel Category=VLMs that take omni inputs2026.06 | 30.63 | 32.89 | 31.28 | 30.36 | 31.13 | |
| Gemini 3.1 FlashModel Category=VLMs that take omni inputs2026.06 | 27.5 | 47.65 | 43.33 | 41.79 | 40.43 | |
| Qwen3 OmniModel Category=Open-source omni models2026.06 | 26.11 | 33.02 | 38.97 | 31.44 | 32.61 | |
| OmniVinciModel Category=Open-source omni models2026.06 | 18.75 | 20.81 | 26.07 | 21.07 | 21.88 | |
| Qwen 3.6 PlusModel Category=VLMs that do not support omni inputs2026.06 | 15 | 26.17 | 14.69 | 19.29 | 18.5 | |
| GPT 5.4Model Category=VLMs that do not support omni inputs2026.06 | 14.37 | 20.13 | 18.96 | 18.57 | 18.12 | |
| GLM 5V TurboModel Category=VLMs that do not support omni inputs2026.06 | 14.37 | 18.12 | 13.27 | 17.5 | 15.88 | |
| Baichuan-Omni-1.5Model Category=Open-source omni models2026.06 | 14.37 | 12.75 | 12.8 | 12.86 | 13.12 |