Skip to content

Benchmark

Arena ELO - Chinese

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

Models measured
340
Scores recorded
396
Sources
1
IQ weight
1.23

Score distribution

1,334.6
Mean
1,384.4
Median
172.2
Std dev
1,067.9
10th percentile
1,512.4
90th percentile
719
Min
7191,628.9
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.23 for chat models, measured over 340 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard

#ModelScore
251Claude Sonnet 3.7Thinking 32K, 20250219· LMArena data1,298.9
252Olmo 3 32B Think· LMArena data1,298.6
253Claude Sonnet 3.720250219· LMArena data1,297.5
254GLM 4 Plus· LMArena data1,296
255Gemini 1.5 Flash002· LMArena data1,294.9
256Grok 22024-08-13· LMArena data1,289.4
257Hunyuan Large Vision· LMArena data1,286.8
258DeepSeek V2.5· LMArena data1,281.4
259Granite 4.2 3B· LMArena data1,278.8
260Llama 4 Maverick· LMArena data1,277.7
261GPT 4o2024-05-13· LMArena data1,277.3
262Llama 3.3 Nemotron Super 49B v1· LMArena data1,276.7
263Gemini 1.5 Pro001· LMArena data1,274.8
264Gemini Advanced0514· LMArena data1,274.1
265Qwen2.5 72B Instruct· LMArena data1,272.5
266Claude 3.5 Sonnet20241022· LMArena data1,271.8
267GPT-4.1 nano2025-04-14· LMArena data1,270.1
268Claude 3.5 Sonnet20240620· LMArena data1,264.8
269GPT-4o mini2024-07-18· LMArena data1,264.6
270Llama 3.1 Nemotron 70B Instruct· LMArena data1,263.1
271Grok 2 mini2024-08-13· LMArena data1,262.5
272Llama 4 Scout· LMArena data1,254.9
273GPT 4o2024-08-06· LMArena data1,254.2
274Qwen Max0919· LMArena data1,254.1
275Mistral Small 3.1 24B Instruct2503· LMArena data1,252.6
276Granite 4.0 H Small· LMArena data1,251.6
277Llama 3.1 Tulu 3 70B· LMArena data1,249.4
278Claude 3 Opus20240229· LMArena data1,247.8
279Nova Pro 1.0· LMArena data1,243.7
280GPT-4 Turbo Preview1106· LMArena data1,242.5
281GPT-4 Turbo2024-04-09· LMArena data1,242.3
282Llama 3.1 405B InstructBF16· LMArena data1,242
283Hunyuan Standard 256K· LMArena data1,241.6
284Reka Core20240904· LMArena data1,241.1
285Olmo 3.1 32B Think· LMArena data1,240.9
286Mistral Large 22407· LMArena data1,240.4
287Qwen2 72B Instruct· LMArena data1,240
288Mistral Large2411· LMArena data1,239
289Athene 70B0725· LMArena data1,237.6
290GPT-4 Turbo Preview0125· LMArena data1,237.6
291Llama 3.1 405B InstructFP8· LMArena data1,234.5
292Gemini 1.5 Flash001· LMArena data1,234.3
293Gemini 1.5 Flash-8B001· LMArena data1,230.9
294Claude Haiku 3.520241022· LMArena data1,228.7
295GLM 40520· LMArena data1,228.6
296Magistral Medium2506· LMArena data1,227.6
297Command R+08-2024· LMArena data1,226.5
298Nova Lite 1.0· LMArena data1,225.3
299Gemma 2 9B IT SimPO· LMArena data1,225.2
300Qwen 2.5 Coder 32B Instruct· LMArena data1,222

Showing 251–300 of 396 results.

Data credits: LMArena, CC BY 4.0 (licence)