Skip to content

Benchmark

Arena ELO Text Factuality - Hard Prompts

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

Models measured
168
Scores recorded
192
Sources
1
IQ weight
2.23

Score distribution

1,459.1
Mean
1,464.1
Median
40.8
Std dev
1,401.2
10th percentile
1,502.4
90th percentile
1,311.6
Min
1,311.61,544.8
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 2.23 for chat models, measured over 168 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard

#ModelScore
51GPT-6.1 TerraMax· LMArena data1,487.9
52Gemini 3.5 FlashMedium· LMArena data1,487
53GLM 5.2Max· LMArena data1,485.2
54Muse Spark 1.1· LMArena data1,484.8
55GPT 5.1High· LMArena data1,483.4
56Gemini 3 Flash· LMArena data1,483.3
57Muse Spark· LMArena data1,482
58Kimi K2.6· LMArena data1,481.6
59Claude Sonnet 4.520250929· LMArena data1,480.5
60Claude Sonnet 4.5High, 32K, 20250929· LMArena data1,480.1
61GLM 5.1· LMArena data1,479.5
62ERNIE 5.1· LMArena data1,479.2
63Grok 4.6High· LMArena data1,478.6
64Claude Sonnet 5High· LMArena data1,478.3
65DeepSeek V4 Pro· LMArena data1,478
66Step 5 Preview· LMArena data1,477.8
67MiMo V2.6 Flash· LMArena data1,476.4
68Grok 4.20Reasoning, 0309, beta· LMArena data1,474.7
69nova-experimental-chat26-02-10· LMArena data1,474.4
70DeepSeek V4 ProHigh, 20260813· LMArena data1,474.4
71Gemini 2.5 Pro· LMArena data1,474.1
72Inkling FP4· LMArena data1,473.7
73MiMo V2 Pro· LMArena data1,473.3
74Grok 4.20 Multi-Agent0309, beta· LMArena data1,472.3
75Qwen3.6 Max Preview· LMArena data1,472.1
76GPT-6 SolMax· LMArena data1,471.9
77Qwen3.7 Plus· LMArena data1,471.4
78GLM 5· LMArena data1,471.3
79Gemma 4 31B IT· LMArena data1,470.7
80GPT-5.4 miniHigh· LMArena data1,470.6
81GPT-5.2 Chat20260210· LMArena data1,470.3
82Beam 1.0 501B A23B· LMArena data1,470.2
83MiMo V2.5· LMArena data1,469.3
84GPT 5.1· LMArena data1,469.1
85DeepSeek-V4-Pro PreviewHigh· LMArena data1,468.4
86Claude Opus 4.120250805· LMArena data1,467.8
87GPT-6.2 AstraMax· LMArena data1,467.6
88Hy3 (Deep Infra)· LMArena data1,467.6
89Qwen3.5 397B A17B· LMArena data1,467.5
90GPT 5.2· LMArena data1,467.2
91Gemini 3.5 Flash Lite· LMArena data1,467
92Kimi K2.5Thinking· LMArena data1,466
93GPT-6.2 SolMax· LMArena data1,464.8
94Kimi K2.5Instant· LMArena data1,464.5
95Grok 4.20 Beta 1· LMArena data1,464.2
96MiMo V2 Omni· LMArena data1,464.2
97GPT-5.5 Instant· LMArena data1,464.1
98Qwen 3.8 27B· LMArena data1,464.1
99GPT 5.2High· LMArena data1,463.8
100Gemini 3 FlashMinimal· LMArena data1,463.6

Showing 51–100 of 192 results.

Data credits: LMArena, CC BY 4.0 (licence)