Skip to content

Benchmark

Arena ELO Vision - Homework

People compare model answers and choose which they prefer. Higher scores are better.

Models measured
104
Scores recorded
119
Sources
1
IQ weight
2.12

Score distribution

1,265.7
Mean
1,276
Median
57.7
Std dev
1,171.2
10th percentile
1,332.9
90th percentile
1,117.1
Min
1,117.11,362.8
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 2.12 for chat models, measured over 104 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard

#ModelScore
101Claude Sonnet 3.7Thinking 32K, 20250219· LMArena data1,192.5
102Gemini 2.5 Flash-Lite PreviewThinking, 06-17· LMArena data1,191.5
103Step3· LMArena data1,183.4
104Gemini 2.0 Flash001· LMArena data1,176.3
105Mistral Small 3.22506· LMArena data1,173.5
106Grok 4.1 FastReasoning· LMArena data1,173.3
107Grok 40709· LMArena data1,171.2
108Mistral Medium 3.12508· LMArena data1,166.7
109Mistral Medium 32505· LMArena data1,165.3
110Llama 4 Maverick· LMArena data1,164.3
111GLM 4.5V· LMArena data1,163.6
112Gemma 3 27B IT· LMArena data1,161.4
113GPT-5 nanoHigh· LMArena data1,159.6
114Mistral Small 3.1 24B Instruct2503· LMArena data1,159.5
115GLM 4.6v· LMArena data1,157.4
116Step-1o Turbo202506· LMArena data1,153
117Llama 4 Scout· LMArena data1,150.4
118Claude 3.5 Sonnet20241022· LMArena data1,135.8
119Claude Haiku 3.520241022· LMArena data1,117.1

Showing 101–119 of 119 results.

Data credits: LMArena, CC BY 4.0 (licence)