Skip to content

Benchmark

MMStar

Vision benchmarks measure how well models understand images. Scores are percentages. Higher is better.

Models measured
68
Scores recorded
72
Sources
3
IQ weight
0.99

Score distribution

65.1%
Mean
67.1%
Median
13.1%
Std dev
47.5%
10th percentile
81.4%
90th percentile
24.9%
Min
24.9%85.9%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 0.99 for chat models, measured over 67 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard

#ModelScore
1Qwen3.8 MaxQwen3.8-Max release vision/video table · published setting· Reported by Qwen· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.85.9%
2Qwen3.5 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.83.8%
3Qwen3.6 Plus· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim83.3%
4Qwen3.6 PlusQwen official release table · thinking mode unless benchmark requires otherwise· Reported by Alibaba Qwen· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.83.3%
5Qwen3.5 122B-A10B· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim82.9%
6Qwen3.5 35B A3B· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim81.9%
7Qwen3.6 27B· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim81.4%
8Qwen3.6 27BQwen official release table · thinking mode unless benchmark requires otherwise· Reported by Alibaba Qwen· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.81.4%
9Qwen3.5 27B· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim81%
10Doubao Seed 1.8· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim79.9%
11Qwen3 VL 32B Thinking· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it.79.4%
12Qwen3.5 Omni PlusQwen3.5-Omni release · published setting· Reported by Qwen, Qwen Team· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.79.4%
13Seed 2.0 Mini· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim79.1%
14Qwen3 VL 235Bthinking· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it.78.7%
15Qwen3 VL 235B· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it.78.4%
16Qwen3 VL 32B Instruct· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it.77.7%
17GPT 5· OpenVLM Leaderboard data75.7%
18Qwen3.5 Omni FlashQwen3.5-Omni release · published setting· Reported by Qwen, Qwen Team· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.75.7%
19qwen3-vl-30bthinking· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it.75.5%
20Qwen: Qwen3 VL 8B Thinking· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it.75.3%
21SenseNova-V6-Pro· OpenVLM Leaderboard data73.7%
22Gemini 2.5 Pro· OpenVLM Leaderboard data73.6%
23InternVL3 78B· OpenVLM Leaderboard data73.4%
24GPT-5 mini· OpenVLM Leaderboard data73.2%
25Hunyuan Standard Vision· OpenVLM Leaderboard data72.9%
26internvl3-38b· OpenVLM Leaderboard data72.6%
27qwen3-vl-30b· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it.72.1%
28Qwen3 VL 8B Instruct· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it.70.9%
29Qwen2.5 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
70.8%
30ChatGPT-4o Latest· OpenVLM Leaderboard data70.2%
31GPT 4.1· OpenVLM Leaderboard data69.8%
32Gemini 2.0 Flash· OpenVLM Leaderboard data69.4%
33GPT-4.5 Preview· OpenVLM Leaderboard data69.3%
34internvl3-8b· OpenVLM Leaderboard data68.7%
35Qwen2-VL-72B· OpenVLM Leaderboard data68.6%
36GPT-5 nano· OpenVLM Leaderboard data68.3%
37gemini-1.5-pro-002· OpenVLM Leaderboard data67.1%
38GLM 4 Vision Plus· OpenVLM Leaderboard data66.3%
39Grok 2 Vision 1212· OpenVLM Leaderboard data65.2%
40gemini-1.5-flash-002· OpenVLM Leaderboard data64.4%
41Qwen2.5-VL 7B Instruct· OpenVLM Leaderboard data
Values from 2 sources
64.1%
42Qwen2.5-Omni 7B· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim64%
43Molmo 72B· OpenVLM Leaderboard data63.3%
44DeepSeek VL2· OpenVLM Leaderboard data
Values from 2 sources
61.9%
45InternVL2 26B· OpenVLM Leaderboard data61%
46GPT-4.1 mini· OpenVLM Leaderboard data60.9%
47Yi Vision· OpenVLM Leaderboard data60.5%
48Moonshot V1 8K· OpenVLM Leaderboard data60.1%
49Gemma 3 27B IT· OpenVLM Leaderboard data59.6%
50Gemini 1.5 Pro· OpenVLM Leaderboard data59.1%

Showing 1–50 of 72 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.