Skip to content

Benchmark

MME

Vision benchmarks measure how well models understand images.

Models measured
41
Scores recorded
41
Sources
2
IQ weight
0.95

Score distribution

2,007.6
Mean
2,148.9
Median
460.7
Std dev
1,586.8
10th percentile
2,422
90th percentile
0.225
Min
0.2252,538.6
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 0.95 for chat models, measured over 41 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard

#ModelScore
1InternVL3 78B· OpenVLM Leaderboard data2,538.6
2Qwen2-VL-72B· OpenVLM Leaderboard data2,504.2
3internvl3-38b· OpenVLM Leaderboard data2,500.7
4Qwen2.5 VL 72B Instruct· OpenVLM Leaderboard data2,435
5internvl3-8b· OpenVLM Leaderboard data2,422
6Gemini 2.0 Flash· OpenVLM Leaderboard data2,386.1
7GPT 4.1· OpenVLM Leaderboard data2,337.6
8Qwen2.5-VL 7B Instruct· OpenVLM Leaderboard data2,312.1
9Qwen-VL Max· OpenVLM Leaderboard data2,281.7
10MiniCPM-V 2.6· OpenVLM Leaderboard data2,268.7
11InternVL2 26B· OpenVLM Leaderboard data2,259.8
12Hunyuan Standard Vision· OpenVLM Leaderboard data2,250.4
13DeepSeek VL2· OpenVLM Leaderboard data
Values from 2 sources
2,230.2
14Qwen-VL Plus· OpenVLM Leaderboard data2,229.8
15Yi Vision· OpenVLM Leaderboard data2,203.7
16qwen2.5-vl-3b-instruct· OpenVLM Leaderboard data2,199.9
17GLM 4 Vision Plus· OpenVLM Leaderboard data2,194.9
18GLM 4 Vision· OpenVLM Leaderboard data2,163.7
19Grok 2 Vision 1212· OpenVLM Leaderboard data2,154.7
20Moonshot V1 8K· OpenVLM Leaderboard data2,151.6
21gemini-1.0-pro· OpenVLM Leaderboard data2,148.9
22Gemma 3 27B IT· OpenVLM Leaderboard data2,148.5
23GPT-4.1 mini· OpenVLM Leaderboard data2,110.9
24Gemini 1.5 Pro· OpenVLM Leaderboard data2,110.6
25Gemini 1.5 Flash· OpenVLM Leaderboard data2,077.9
26InternVL2 4B· OpenVLM Leaderboard data2,064.6
27Gemma 3 12B IT· OpenVLM Leaderboard data2,053.5
28Molmo 72B· OpenVLM Leaderboard data1,992
29Phi 4 Multimodal Instruct· OpenVLM Leaderboard data1,961.7
30Pixtral 12B· OpenVLM Leaderboard data1,921.7
31Molmo 7B-D· OpenVLM Leaderboard data1,853.2
32Phi-3.5 Vision Instruct· OpenVLM Leaderboard data1,838.1
33Llama 3.2 11B Vision Instruct· OpenVLM Leaderboard data1,820.5
34GPT-4.1 nano· OpenVLM Leaderboard data1,789.6
35Gemma 3 4B· OpenVLM Leaderboard data1,744.1
36Llama 3.2 90B Vision Instruct· OpenVLM Leaderboard data1,741
37Claude 3 Sonnet· OpenVLM Leaderboard data1,625.9
38Claude 3 Opus· OpenVLM Leaderboard data1,586.8
39Moondream2· OpenVLM Leaderboard data1,528.3
40Claude 3 Haiku· OpenVLM Leaderboard data1,453.2
41kosmos-2· OpenVLM Leaderboard data721.1

Data credits: OpenVLM Leaderboard (OpenCompass / VLMEvalKit), Apache-2.0 (licence) · Data by LLM Stats