Skip to content

Benchmark

Arena ELO - Japanese

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

Models measured
245
Scores recorded
286
Sources
1
IQ weight
1.65

Score distribution

1,296.5
Mean
1,336.3
Median
150
Std dev
1,061.5
10th percentile
1,461.1
90th percentile
878.3
Min
878.31,557.4
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.65 for chat models, measured over 245 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 3 FlashMinimal· LMArena data1,427.4
52Kimi K2.5Thinking· LMArena data1,425.4
53GPT 5.1· LMArena data1,422.6
54Claude Sonnet 5High· LMArena data1,422.4
55ERNIE 5.1· LMArena data1,422
56GPT-6 AstraMax· LMArena data1,421.2
57Claude Sonnet 4.6· LMArena data1,420.3
58GPT 5.2High· LMArena data1,419.5
59GPT-5.6 LunaxHigh· LMArena data1,417.8
60Grok 4.20Reasoning, 0309, beta· LMArena data1,416.8
61Claude Opus 4.5High, 32K, 20251101· LMArena data1,416.7
62GPT-4.5 Preview2025-02-27· LMArena data1,416.1
63GLM 4.5· LMArena data1,415.8
64GLM 5· LMArena data1,414.8
65DeepSeek-V4-Pro PreviewHigh· LMArena data1,413.6
66GPT-5.2 Chat20260210· LMArena data1,413.4
67MiMo V2.5 Pro· LMArena data1,410.5
68Claude Opus 4.520251101· LMArena data1,410.4
69Qwen3.7 Plus· LMArena data1,409.9
70Gemini 3.1 Flash Lite· LMArena data1,409.3
71GPT-5 Chat· LMArena data1,409.2
72GPT-5.5 Instant· LMArena data1,408.2
73Doubao Seed 2.0 Pro· LMArena data1,406.4
74Grok 4.20 Multi-Agent0309, beta· LMArena data1,405.8
75ChatGPT-4o Latest20250326· LMArena data1,405.2
76Gemini 2.5 Flash Preview09-2025· LMArena data1,405
77DeepSeek-V4-Flash PreviewHigh· LMArena data1,404.5
78OpenAI o32025-04-16· LMArena data1,403.6
79GPT 5High· LMArena data1,401.9
80Qwen3 235B A22B Instruct 25072507· LMArena data1,400
81Hy3 (Deep Infra)· LMArena data1,399.5
82Inkling Small· LMArena data1,399.3
83Qwen3 Max Preview· LMArena data1,398.5
84Step 5 Preview· LMArena data1,397.9
85Grok 4.1· LMArena data1,397.9
86Mistral Large 3· LMArena data1,397.2
87Gemini 2.5 Flash· LMArena data1,396.8
88Qwen3 Next 80B A3B Instruct· LMArena data1,395.6
89GLM 4.6· LMArena data1,393.9
90Grok 40709· LMArena data1,393.4
91Qwen3.6 Plus· LMArena data1,391.7
92DeepSeek R10528· LMArena data1,391.2
93Qwen 3.8 27B· LMArena data1,389.7
94Claude Sonnet 4.520250929· LMArena data1,389.5
95GPT 5.2· LMArena data1,389.3
96DeepSeek V4 Flash· LMArena data1,387.3
97Grok 3 Preview02-24· LMArena data1,386.7
98Qwen 3 235B A22B2507, thinking· LMArena data1,386.7
99Qwen3 VL 235B· LMArena data1,385.3
100Mistral Medium 3.12508· LMArena data1,380.2

Showing 51–100 of 286 results.

Data credits: LMArena, CC BY 4.0 (licence)