HallusionBench — leaderboard
HallusionBench evaluates multimodal large language models on visual illusion and hallucination-style image-text reasoning cases.
Metric: Overall Score. Source: huggingface.co. Status: saturation imminent. 284 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5 (2025-08-07) | 65.2 |
| 2 | Gemini 2.5 Pro | 64.1 |
| 3 | GPT-5 Mini (2025-08-07) | 62.5 |
| 4 | GPT-5 Nano | 60.9 |
| 5 | GPT-4.5 | 60 |
| 6 | InternVL3-78B | 59.1 |
| 7 | Qwen 2 VL 72B | 58.7 |
| 8 | GPT-4.1 (2025-04-14) | 58.5 |
| 9 | Doubao-1.5-Pro | 58.5 |
| 10 | InternVL3-38B | 58.4 |
| 11 | Gemini 2.0 Flash | 58 |
| 12 | InternVL2.5-78B | 57.4 |
| 13 | GPT-4o ChatGPT | 57 |
| 14 | Ovis2-8B | 56.3 |
| 15 | InternVL3-14B | 55.9 |
Interactive version: theaggregate.ai/benchmark?slug=hallusionbench · How the rankings work · Data refreshed daily, snapshot 2026-07-22.