Skip to content

Benchmark

MMBench V1.1 CN

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

Models measured
50
Scores recorded
50
Sources
1
IQ weight
1.14

Score distribution

74.3%
Mean
78%
Median
14.8%
Std dev
62.8%
10th percentile
86.4%
90th percentile
0.9%
Min
0.9%88.1%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.14 for chat models, measured over 50 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard

#ModelScore
1Gemini 2.5 Pro· OpenVLM Leaderboard data88.1%
2InternVL3 78B· OpenVLM Leaderboard data87.9%
3SenseNova-V6-Pro· OpenVLM Leaderboard data87.6%
4Qwen2.5 VL 72B Instruct· OpenVLM Leaderboard data87.4%
5internvl3-38b· OpenVLM Leaderboard data86.6%
6GPT 5· OpenVLM Leaderboard data86.4%
7GPT 4.1· OpenVLM Leaderboard data86.2%
8GPT-5 mini· OpenVLM Leaderboard data86.1%
9Qwen2-VL-72B· OpenVLM Leaderboard data85.8%
10ChatGPT-4o Latest· OpenVLM Leaderboard data85.2%
11GPT-4.5 Preview· OpenVLM Leaderboard data83.3%
12Hunyuan Standard Vision· OpenVLM Leaderboard data83.3%
13Grok 2 Vision 1212· OpenVLM Leaderboard data82.5%
14Qwen2.5-VL 7B Instruct· OpenVLM Leaderboard data82.1%
15internvl3-8b· OpenVLM Leaderboard data81.9%
16gemini-1.5-pro-002· OpenVLM Leaderboard data81.1%
17gemini-1.5-flash-002· OpenVLM Leaderboard data81%
18GLM 4 Vision Plus· OpenVLM Leaderboard data81%
19InternVL2 26B· OpenVLM Leaderboard data80.9%
20DeepSeek VL2· OpenVLM Leaderboard data80.1%
21GPT-5 nano· OpenVLM Leaderboard data79.6%
22Molmo 72B· OpenVLM Leaderboard data79.6%
23GPT-4.1 mini· OpenVLM Leaderboard data78.6%
24Gemma 3 27B IT· OpenVLM Leaderboard data78.2%
25GLM 4 Vision· OpenVLM Leaderboard data78%
26MiniCPM-V 2.6· OpenVLM Leaderboard data76.9%
27Gemini 1.5 Flash· OpenVLM Leaderboard data76.7%
28qwen2.5-vl-3b-instruct· OpenVLM Leaderboard data76.5%
29Moonshot V1 8K· OpenVLM Leaderboard data76.2%
30Yi Vision· OpenVLM Leaderboard data76.1%
31Llama 3.2 90B Vision Instruct· OpenVLM Leaderboard data75.2%
32Phi 4 Multimodal Instruct· OpenVLM Leaderboard data74.9%
33Gemma 3 12B IT· OpenVLM Leaderboard data74.3%
34Qwen-VL Max· OpenVLM Leaderboard data73.7%
35Gemini 1.5 Pro· OpenVLM Leaderboard data73.2%
36Gemini 2.0 Flash· OpenVLM Leaderboard data71.6%
37Molmo 7B-D· OpenVLM Leaderboard data71.6%
38InternVL2 4B· OpenVLM Leaderboard data71.4%
39Pixtral 12B· OpenVLM Leaderboard data70.1%
40gemini-1.0-pro· OpenVLM Leaderboard data69.2%
41Qwen-VL Plus· OpenVLM Leaderboard data68.2%
42Gemma 3 4B· OpenVLM Leaderboard data66.8%
43Llama 3.2 11B Vision Instruct· OpenVLM Leaderboard data64.2%
44Claude 3 Sonnet· OpenVLM Leaderboard data63.5%
45Phi-3.5 Vision Instruct· OpenVLM Leaderboard data62.8%
46GPT-4.1 nano· OpenVLM Leaderboard data60.4%
47Claude 3 Opus· OpenVLM Leaderboard data58.1%
48Claude 3 Haiku· OpenVLM Leaderboard data56.2%
49Moondream2· OpenVLM Leaderboard data28.5%
50kosmos-2· OpenVLM Leaderboard data0.9%

Data credits: OpenVLM Leaderboard (OpenCompass / VLMEvalKit), Apache-2.0 (licence)