OpenVLM HallusionBench - qAcc: leaderboard
Metric: Accuracy (%). Source: huggingface.co. 284 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 2.5 Pro | 56.3 |
| 2 | GPT-5 | 56 |
| 3 | GPT-5 Mini (2025-08-07) | 53 |
| 4 | GPT-5 Nano | 52.7 |
| 5 | Gemini 2.0 Flash | 51.4 |
| 6 | InternVL3-78B | 50.8 |
| 7 | Qwen 2 VL 72B | 49.7 |
| 8 | InternVL3-38B | 49.7 |
| 9 | GPT-4.5 | 49.2 |
| 10 | Gemini 1.5 Pro (002) | 49 |
| 11 | InternVL2.5-78B | 48.1 |
| 12 | GPT-4.1 (2025-04-14) | 48.1 |
| 13 | GPT-4o ChatGPT | 47.3 |
| 14 | Gemini 1.5 Flash (002) | 47 |
| 15 | Claude 3.7 Sonnet | 46.8 |
Interactive version: theaggregate.ai/benchmark?slug=openvlm-hallusionbench-qacc · How It Works · Data refreshed daily, snapshot 2026-09-05.