Skip to content

Benchmark

Arena ELO Text Factuality - Polish

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

Models measured
57
Scores recorded
70
Sources
1
IQ weight
1.96

Score distribution

1,431.3
Mean
1,432.8
Median
31.9
Std dev
1,391.4
10th percentile
1,473.6
90th percentile
1,352.1
Min
1,352.11,488.1
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.96 for chat models, measured over 57 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard

#ModelScore
51Gemini 2.5 Flash· LMArena data1,408.4
52DeepSeek V3.2· LMArena data1,406.7
53Claude Sonnet 4.520250929· LMArena data1,406.1
54MiMo V2 FlashNon-Thinking· LMArena data1,402.4
55DeepSeek-V4-Flash PreviewHigh· LMArena data1,402.2
56Claude Sonnet 4.5High, 32K, 20250929· LMArena data1,401.8
57Mistral Medium 3.12508· LMArena data1,401.4
58DeepSeek V3.2Thinking· LMArena data1,399
59Grok 4.3· LMArena data1,396.2
60Qwen3 235B A22B Instruct 25072507· LMArena data1,395.6
61Grok 4.1 FastReasoning· LMArena data1,394.4
62Step 3.5 Flash· LMArena data1,393.4
63Claude Opus 4.120250805· LMArena data1,391.4
64MiniMax M2.7· LMArena data1,388.7
65GPT-6 LunaMax· LMArena data1,388.4
66Kimi K2 Thinking· LMArena data1,378.8
67Gemini 2.5 Flash-Lite PreviewNon-Thinking, 09-2025· LMArena data1,373.7
68Qwen3.5 Flash· LMArena data1,372.5
69Claude Haiku 4.520251001· LMArena data1,357.8
70MiniMax M2.5· LMArena data1,352.1

Showing 51–70 of 70 results.

Data credits: LMArena, CC BY 4.0 (licence)