Benchmark
HallusionBench
Vision benchmarks measure how well models understand images. Scores are percentages. Higher is better.
- Models measured
- 64
- Scores recorded
- 66
- Sources
- 3
- IQ weight
- 0.90
Score distribution
- 53.3%
- Mean
- 53.9%
- Median
- 10.7%
- Std dev
- 40.5%
- 10th percentile
- 67.1%
- 90th percentile
- 19.8%
- Min
This benchmark has a weight of 0.90 for chat models, measured over 64 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.
Leaderboard
| # | Model | Score |
|---|---|---|
| 1 | Qwen3.5 397B A17BQwen3.5 release/model card · published setting· Reported by Qwen· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 71.4% |
| 2 | Qwen3.5 27B· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 70% |
| 3 | Qwen3.6 35B A3B· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 69.8% |
| 4 | Qwen3.6 FlashQwen3.6-35B-A3B official model card · hosted API name Qwen3.6-Flash· Reported by Alibaba Qwen· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 69.8% |
| 5 | Qwen3.5 35B A3B· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 67.9% |
| 6 | Qwen3.5 122B-A10B· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 67.6% |
| 7 | Qwen3 VL 32B Thinking· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. | 67.4% |
| 8 | SenseNova-V6-Pro· OpenVLM Leaderboard data | 67.1% |
| 9 | Qwen3 VL 235Bthinking· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. | 66.7% |
| 10 | qwen3-vl-30bthinking· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. | 66% |
| 11 | Qwen: Qwen3 VL 8B Thinking· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. | 65.4% |
| 12 | GPT 5· OpenVLM Leaderboard data | 65.2% |
| 13 | Seed 2.0 Mini· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 65.1% |
| 14 | Gemini 2.5 Pro· OpenVLM Leaderboard data | 64.1% |
| 15 | Doubao Seed 1.8· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 63.9% |
| 16 | Qwen3 VL 32B Instruct· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. | 63.8% |
| 17 | Qwen3 VL 235B· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. | 63.2% |
| 18 | GPT-5 mini· OpenVLM Leaderboard data | 62.5% |
| 19 | qwen3-vl-30b· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. | 61.5% |
| 20 | Qwen3 VL 8B Instruct· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. | 61.1% |
| 21 | GPT-5 nano· OpenVLM Leaderboard data | 60.9% |
| 22 | GPT-4.5 Preview· OpenVLM Leaderboard data | 60% |
| 23 | InternVL3 78B· OpenVLM Leaderboard data | 59.1% |
| 24 | Qwen2-VL-72B· OpenVLM Leaderboard data | 58.7% |
| 25 | GPT 4.1· OpenVLM Leaderboard data | 58.5% |
| 26 | internvl3-38b· OpenVLM Leaderboard data | 58.4% |
| 27 | Gemini 2.0 Flash· OpenVLM Leaderboard data | 58% |
| 28 | Hunyuan Standard Vision· OpenVLM Leaderboard data | 57.7% |
| 29 | ChatGPT-4o Latest· OpenVLM Leaderboard data | 57% |
| 30 | gemini-1.5-pro-002· OpenVLM Leaderboard data | 55.9% |
| 31 | GLM 4 Vision Plus· OpenVLM Leaderboard data | 55.6% |
| 32 | Qwen2.5 VL 72B Instruct· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claimValues from 2 sources
| 55.2% |
| 33 | gemini-1.5-flash-002· OpenVLM Leaderboard data | 53.9% |
| 34 | Qwen2.5-VL 7B Instruct· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claimValues from 2 sources
| 52.9% |
| 35 | Grok 2 Vision 1212· OpenVLM Leaderboard data | 51.7% |
| 36 | InternVL2 26B· OpenVLM Leaderboard data | 51.5% |
| 37 | Yi Vision· OpenVLM Leaderboard data | 50.3% |
| 38 | GPT-4.1 mini· OpenVLM Leaderboard data | 49.3% |
| 39 | internvl3-8b· OpenVLM Leaderboard data | 49% |
| 40 | Gemma 3 27B IT· OpenVLM Leaderboard data | 48.8% |
| 41 | Gemini 1.5 Flash· OpenVLM Leaderboard data | 48.5% |
| 42 | MiniCPM-V 2.6· OpenVLM Leaderboard data | 48.1% |
| 43 | Gemma 3 12B IT· OpenVLM Leaderboard data | 47.2% |
| 44 | Pixtral 12B· OpenVLM Leaderboard data | 47% |
| 45 | Molmo 72B· OpenVLM Leaderboard data | 46.6% |
| 46 | qwen2.5-vl-3b-instruct· OpenVLM Leaderboard data | 46.6% |
| 47 | Molmo 7B-D· OpenVLM Leaderboard data | 46.4% |
| 48 | Moonshot V1 8K· OpenVLM Leaderboard data | 46.1% |
| 49 | gemini-1.0-pro· OpenVLM Leaderboard data | 45.7% |
| 50 | Gemini 1.5 Pro· OpenVLM Leaderboard data | 45.6% |
Showing 1–50 of 66 results.
Data credits: OpenVLM Leaderboard (OpenCompass / VLMEvalKit), Apache-2.0 (licence) · Data by LLM Stats · BenchBoard by SYuan03 (github.com/SYuan03/benchboard). MIT covers its code and data organization; no results licence is inferred. Numeric relays retain the original stating-source links.