OpenVLM Q-Bench - Yes-or-No (Distortion): leaderboard
Metric: Accuracy (%). Source: huggingface.co. 160 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Qwen 2 VL 72B | 91 |
| 2 | Qwen 2 VL 7B | 89.4 |
| 3 | GPT-4.1 Mini | 88.8 |
| 4 | GPT-4.1 (2025-04-14) | 86.2 |
| 5 | InternVL3-78B | 85.6 |
| 6 | Qwen 2 VL 2B | 85.6 |
| 7 | Moonshot v1 8K | 85.1 |
| 8 | InternVL3-8B | 83.5 |
| 9 | Aquila-VL-2B | 82.4 |
| 10 | Gemini 1.5 Pro | 81.9 |
| 11 | GPT-4.1 Nano | 81.4 |
| 12 | InternVL3-14B | 81.4 |
| 13 | InternVL2-8B | 79.8 |
| 14 | Grok 2 (1212) | 79.8 |
| 15 | Pixtral-12B | 78.2 |
Interactive version: theaggregate.ai/benchmark?slug=openvlm-q-bench-yes-or-no-distortion · How It Works · Data refreshed daily, snapshot 2026-09-19.