OpenVLM MMBench V1.0 CN - RR: leaderboard
MMBench v1.0 Chinese test questions on relation reasoning (RR: social, physical and natural relations), from the OpenCompass OpenVLM Leaderboard (VLMEvalKit). CircularEval counts a question only if every rotation of its options is answered correctly. Accuracy in percent, higher is better; this is the original question set, separate from the MMBench V1.1 boards.
Metric: Accuracy (%). Source: huggingface.co. Saturation forecast: Estimated already saturated. 230 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | InternVL2.5-78B | 89.1 | #434 |
| 2 | InternVL3-78B | 87.7 | #361 |
| 3 | InternVL3-38B | 84.4 | #413 |
| 4 | InternVL3-14B | 83.4 | #506 |
| 5 | InternVL3-8B | 82.5 | #624 |
| 6 | GPT-4.1 (2025-04-14) | 81.5 | #257 |
| 7 | GPT-4.1 Mini | 80.6 | #363 |
| 8 | Qwen 2 VL 72B | 80.1 | #450 |
| 9 | Grok 2 (1212) | 78.7 | #502 |
| 10 | Qwen 2 VL 7B | 77.7 | #743 |
| 11 | Gemini 1.0 Pro | 77.7 | #900 |
| 12 | Claude 3.5 Sonnet (20241022) | 77.3 | #302 |
| 13 | Aquila-VL-2B | 76.8 | #1216 |
| 14 | Gemma 3 27B | 76.3 | #606 |
| 15 | Gemma 3 12B | 76.3 | #674 |
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-v1-0-cn-rr · How It Works · Data refreshed daily, snapshot 2026-11-05.