Skip to content

Benchmark

Arena ELO Text Factuality

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

Models measured
170
Scores recorded
194
Sources
1
IQ weight
2.26

Score distribution

1,435.9
Mean
1,441.9
Median
39.3
Std dev
1,379.4
10th percentile
1,478.9
90th percentile
1,299.6
Min
1,299.61,522.2
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 2.26 for chat models, measured over 170 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 Pro· LMArena data1,461.8
52GLM 5.3 Flash· LMArena data1,461.2
53Claude Sonnet 4.6· LMArena data1,460.8
54GPT-5.6 LunaxHigh· LMArena data1,460
55Claude Opus 4.520251101· LMArena data1,459.3
56Claude Opus 4.5High, 32K, 20251101· LMArena data1,459
57GPT 5.1High· LMArena data1,458.1
58Grok 4.20 Multi-Agent0309, beta· LMArena data1,457.6
59ERNIE 5.1· LMArena data1,456.5
60Kimi K2.6· LMArena data1,456.4
61GLM 5.1· LMArena data1,456.2
62Grok 4.20Reasoning, 0309, beta· LMArena data1,456
63GPT-6.1 TerraMax· LMArena data1,455.4
64DeepSeek V4 Pro· LMArena data1,455
65DeepSeek V4 ProHigh, 20260813· LMArena data1,452.9
66Grok 4.20 Beta 1· LMArena data1,452.6
67Qwen3.6 Max Preview· LMArena data1,452.3
68Qwen3.7 Plus· LMArena data1,452.3
69GLM 5· LMArena data1,452
70Inkling FP4· LMArena data1,451.8
71nova-experimental-chat26-02-10· LMArena data1,451.8
72Claude Sonnet 5High· LMArena data1,451.4
73Grok 4.6High· LMArena data1,451.3
74Gemini 3 FlashMinimal· LMArena data1,451.2
75GPT-5.2 Chat20260210· LMArena data1,450.4
76MiMo V2.6 Flash· LMArena data1,449.8
77Gemma 4 31B IT· LMArena data1,449.1
78Step 5 Preview· LMArena data1,448.9
79Claude Sonnet 4.520250929· LMArena data1,448.7
80DeepSeek-V4-Pro PreviewHigh· LMArena data1,447.6
81Hy3 (Deep Infra)· LMArena data1,447.3
82GLM 4.6· LMArena data1,447.1
83DeepSeek V3.2 ExpThinking· LMArena data1,446.7
84Gemini 3.5 Flash Lite· LMArena data1,446.6
85GPT-5.4 miniHigh· LMArena data1,446.4
86Claude Sonnet 4.5High, 32K, 20250929· LMArena data1,446.1
87ERNIE 5.00110· LMArena data1,445.3
88Kimi K2.5Thinking· LMArena data1,444.9
89Gemma 4 26B A4B· LMArena data1,444.5
90Qwen3.5 397B A17B· LMArena data1,444.4
91GLM 4.7· LMArena data1,443.9
92MiMo V2 Pro· LMArena data1,443.9
93GPT 5.1· LMArena data1,443.9
94ChatGPT-4o Latest20250326· LMArena data1,443.8
95GPT-6.2 SolMax· LMArena data1,443.3
96GPT-5.5 Instant· LMArena data1,442.9
97ERNIE-5.0-Preview1203· LMArena data1,442.1
98Qwen3 Max Preview· LMArena data1,441.9
99GPT-6 SolMax· LMArena data1,441.8
100GPT 5.2· LMArena data1,441.3

Showing 51–100 of 194 results.

Data credits: LMArena, CC BY 4.0 (licence)