Skip to content

Benchmark

Arena ELO Text Factuality - Industry Mathematical

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

Models measured
102
Scores recorded
117
Sources
1
IQ weight
2.25

Score distribution

1,453.4
Mean
1,452.4
Median
30.2
Std dev
1,416.5
10th percentile
1,489.5
90th percentile
1,363.5
Min
1,363.51,520.2
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 2.25 for chat models, measured over 102 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· LMArena data1,420.7
102nova-experimental-chat11-10· LMArena data1,420.4
103GPT-5.5 Instant· LMArena data1,417.8
104Qwen3.5 35B A3B· LMArena data1,417.1
105Grok 4.1Thinking· LMArena data1,416.5
106ChatGPT-4o Latest20250326· LMArena data1,416.3
107Qwen3.5 Flash· LMArena data1,412.7
108Doubao Seed 2.0 Pro· LMArena data1,411.8
109Grok 4.3· LMArena data1,411.2
110Grok 4.1 FastReasoning· LMArena data1,402.9
111GPT-5.3 Chat· LMArena data1,402.4
112Grok 4.1· LMArena data1,401.4
113MiMo V2 FlashNon-Thinking· LMArena data1,400.8
114Gemini 2.5 Flash-Lite PreviewNon-Thinking, 09-2025· LMArena data1,399.2
115MiniMax M2.5· LMArena data1,389
116Nemotron 3.5 Lightning 30B A3BNVFP4· LMArena data1,387.4
117Trinity Large Preview· LMArena data1,363.5

Showing 101–117 of 117 results.

Data credits: LMArena, CC BY 4.0 (licence)