OpenVLM Q-Bench - Yes-or-No (In-context Distortion): leaderboard
Q-Bench accuracy on yes-or-no questions about distortions of a particular object or region, from the OpenCompass OpenVLM Leaderboard (VLMEvalKit) on the Q-Bench dev split. Q-Bench probes low-level visual perception; each of its twelve question-type by concern cells holds 70-188 of the 1,495 dev questions. Accuracy in percent, higher is better.
Metric: Accuracy (%). Source: huggingface.co. Saturation forecast: Estimated already saturated. 160 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 2.5 VL 7B | 82.9 | #641 |
| 2 | InternVL3-38B | 82.1 | #413 |
| 3 | GPT-4.1 (2025-04-14) | 82.1 | #257 |
| 4 | Qwen 2 VL 72B | 82.1 | #450 |
| 5 | Moonshot v1 8K | 80.3 | #622 |
| 6 | GPT-4.1 Mini | 79.5 | #363 |
| 7 | Gemini 1.5 Pro | 77.8 | #454 |
| 8 | InternVL3-14B | 76.1 | #506 |
| 9 | Gemini 1.5 Flash | 76.1 | #620 |
| 10 | GPT-4.1 Nano | 75.2 | #730 |
| 11 | Qwen 2 VL 7B | 75.2 | #743 |
| 12 | Qwen 2 VL 2B | 75.2 | #1191 |
| 13 | InternVL3-78B | 74.4 | #361 |
| 14 | MiniCPM-V-2.6 | 73.5 | #827 |
| 15 | InternVL3-8B | 71.8 | #624 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=openvlm-q-bench-yes-or-no-in-context-distortion · How It Works · Data refreshed daily, snapshot 2026-11-05.