Skip to content

Benchmark

MathVista

Math benchmarks measure how well models solve maths problems. Scores are percentages. Higher is better.

Models measured
71
Scores recorded
73
Sources
3
IQ weight
1.22

Score distribution

62.7%
Mean
66.1%
Median
14.8%
Std dev
45%
10th percentile
79.2%
90th percentile
0%
Min
0%90.7%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.22 for chat models, measured over 70 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard

#ModelScore
51Pixtral 12B2409· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim58%
52Yi Vision· OpenVLM Leaderboard data57.8%
53GPT-4o mini· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim56.7%
54Pixtral 12B· OpenVLM Leaderboard data56.3%
55GPT-4.1 nano· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim
Values from 2 sources
56.2%
56Gemma 3 12B IT· OpenVLM Leaderboard data56.1%
57Molmo 72B· OpenVLM Leaderboard data55.2%
58Gemini 1.5 Flash-8B· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim54.7%
59Llama 3.2 11B Instruct· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim51.5%
60Molmo 7B-D· OpenVLM Leaderboard data48.7%
61Llama 3.2 11B Vision Instruct· OpenVLM Leaderboard data47.7%
62gemini-1.0-pro· LLM Stats data
Values from 2 sources
46.6%
63Gemma 3 4B· OpenVLM Leaderboard data46.3%
64Claude 3 Opus· OpenVLM Leaderboard data46.1%
65GLM 4 Vision· OpenVLM Leaderboard data45.9%
66Claude 3 Sonnet· OpenVLM Leaderboard data45%
67Phi-3.5 Vision Instruct· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim
Values from 2 sources
43.9%
68Qwen-VL Max· OpenVLM Leaderboard data43.6%
69Claude 3 Haiku· OpenVLM Leaderboard data42.6%
70Qwen-VL Plus· OpenVLM Leaderboard data38.2%
71Moondream2· OpenVLM Leaderboard data33.8%
72kosmos-2· OpenVLM Leaderboard data24.5%
73GPT-3.5 Turbo· LLM Stats data0%

Showing 51–73 of 73 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.