OpenVLM MMBench V1.1 CN - Social Relation: leaderboard
OpenCompass OpenVLM MMBench V1.1 Chinese split accuracy for social relation reasoning.
Metric: Accuracy (%). Source: huggingface.co. Saturation forecast: Estimated already saturated. 284 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | InternVL2.5-78B | 91.2 | #431 |
| 2 | InternVL2-8B | 90.1 | #817 |
| 3 | Qwen 2.5 VL 7B | 90.1 | #643 |
| 4 | InternVL3-8B | 89 | #620 |
| 5 | Gemini 1.5 Flash (002) | 89 | #583 |
| 6 | InternVL3-38B | 89 | #405 |
| 7 | Qwen 2 VL 7B | 89 | #750 |
| 8 | Qwen 2 VL 72B | 89 | #460 |
| 9 | Gemma 3 12B | 89 | #671 |
| 10 | GPT-4o ChatGPT | 89 | |
| 11 | GPT-4.1 (2025-04-14) | 87.9 | #250 |
| 12 | InternVL3-14B | 87.9 | #505 |
| 13 | Gemma 3 4B | 87.9 | #1089 |
| 14 | Gemini 2.5 Pro | 86.8 | #152 |
| 15 | GPT-5 | 86.8 | #98 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=openvlm-mmbench-v1-1-cn-social-relation · How It Works · Data refreshed daily, snapshot 2026-10-20.