Skip to content

Benchmark

Arena ELO Vision - Diagram

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

Models measured
108
Scores recorded
124
Sources
1
IQ weight
1.87

Score distribution

1,253.6
Mean
1,264.9
Median
64
Std dev
1,152.7
10th percentile
1,322.5
90th percentile
1,074
Min
1,0741,354.2
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.87 for chat models, measured over 108 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard

#ModelScore
101Gemini 2.5 Flash-Lite PreviewThinking, 06-17· LMArena data1,185.8
102Step-1o Turbo202506· LMArena data1,185.2
103GLM 4.6v· LMArena data1,181.5
104GPT-4.1 mini2025-04-14· LMArena data1,181.3
105Claude Opus 4Thinking 16K, 20250514· LMArena data1,178.6
106Claude Sonnet 420250514· LMArena data1,176.3
107Claude Opus 420250514· LMArena data1,176.3
108Mistral Medium 3.12508· LMArena data1,174.6
109Claude Sonnet 3.7Thinking 32K, 20250219· LMArena data1,173.9
110Mistral Medium 32505· LMArena data1,155.1
111Gemma 3 27B IT· LMArena data1,154.7
112Claude Sonnet 3.720250219· LMArena data1,152.7
113Step3· LMArena data1,152.3
114GLM 4.5V· LMArena data1,150.2
115Gemini 2.0 Flash001· LMArena data1,148.5
116GPT-5 nanoHigh· LMArena data1,138.9
117Hunyuan Large Vision· LMArena data1,134.6
118Claude 3.5 Sonnet20241022· LMArena data1,133.1
119Llama 4 Maverick· LMArena data1,132.9
120Mistral Small 3.22506· LMArena data1,125.7
121Llama 4 Scout· LMArena data1,125.5
122Mistral Small 3.1 24B Instruct2503· LMArena data1,119.5
123Claude Haiku 3.520241022· LMArena data1,075.2
124Molmo 2 8B· LMArena data1,074

Showing 101–124 of 124 results.

Data credits: LMArena, CC BY 4.0 (licence)