Skip to content

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
19.8%71.4%
The band covers the middle 80% of scores. The line marks the median.

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

#ModelScore
1Qwen3.5 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%
2Qwen3.5 27B· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim70%
3Qwen3.6 35B A3B· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim69.8%
4Qwen3.6 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%
5Qwen3.5 35B A3B· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim67.9%
6Qwen3.5 122B-A10B· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim67.6%
7Qwen3 VL 32B Thinking· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it.67.4%
8SenseNova-V6-Pro· OpenVLM Leaderboard data67.1%
9Qwen3 VL 235Bthinking· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it.66.7%
10qwen3-vl-30bthinking· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it.66%
11Qwen: Qwen3 VL 8B Thinking· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it.65.4%
12GPT 5· OpenVLM Leaderboard data65.2%
13Seed 2.0 Mini· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim65.1%
14Gemini 2.5 Pro· OpenVLM Leaderboard data64.1%
15Doubao Seed 1.8· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim63.9%
16Qwen3 VL 32B Instruct· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it.63.8%
17Qwen3 VL 235B· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it.63.2%
18GPT-5 mini· OpenVLM Leaderboard data62.5%
19qwen3-vl-30b· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it.61.5%
20Qwen3 VL 8B Instruct· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it.61.1%
21GPT-5 nano· OpenVLM Leaderboard data60.9%
22GPT-4.5 Preview· OpenVLM Leaderboard data60%
23InternVL3 78B· OpenVLM Leaderboard data59.1%
24Qwen2-VL-72B· OpenVLM Leaderboard data58.7%
25GPT 4.1· OpenVLM Leaderboard data58.5%
26internvl3-38b· OpenVLM Leaderboard data58.4%
27Gemini 2.0 Flash· OpenVLM Leaderboard data58%
28Hunyuan Standard Vision· OpenVLM Leaderboard data57.7%
29ChatGPT-4o Latest· OpenVLM Leaderboard data57%
30gemini-1.5-pro-002· OpenVLM Leaderboard data55.9%
31GLM 4 Vision Plus· OpenVLM Leaderboard data55.6%
32Qwen2.5 VL 72B Instruct· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim
Values from 2 sources
55.2%
33gemini-1.5-flash-002· OpenVLM Leaderboard data53.9%
34Qwen2.5-VL 7B Instruct· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim
Values from 2 sources
52.9%
35Grok 2 Vision 1212· OpenVLM Leaderboard data51.7%
36InternVL2 26B· OpenVLM Leaderboard data51.5%
37Yi Vision· OpenVLM Leaderboard data50.3%
38GPT-4.1 mini· OpenVLM Leaderboard data49.3%
39internvl3-8b· OpenVLM Leaderboard data49%
40Gemma 3 27B IT· OpenVLM Leaderboard data48.8%
41Gemini 1.5 Flash· OpenVLM Leaderboard data48.5%
42MiniCPM-V 2.6· OpenVLM Leaderboard data48.1%
43Gemma 3 12B IT· OpenVLM Leaderboard data47.2%
44Pixtral 12B· OpenVLM Leaderboard data47%
45Molmo 72B· OpenVLM Leaderboard data46.6%
46qwen2.5-vl-3b-instruct· OpenVLM Leaderboard data46.6%
47Molmo 7B-D· OpenVLM Leaderboard data46.4%
48Moonshot V1 8K· OpenVLM Leaderboard data46.1%
49gemini-1.0-pro· OpenVLM Leaderboard data45.7%
50Gemini 1.5 Pro· OpenVLM Leaderboard data45.6%

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.