Skip to content

Benchmark

Arena ELO - Expert

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

Models measured
318
Scores recorded
371
Sources
1
IQ weight
1.30

Score distribution

1,333.9
Mean
1,371.6
Median
146.5
Std dev
1,113
10th percentile
1,491.8
90th percentile
953.7
Min
953.71,581.3
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.30 for chat models, measured over 318 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard

#ModelScore
251Athene V2 Chat· LMArena data1,268
252DeepSeek V2.51210· LMArena data1,266
253Claude 3.5 Sonnet20241022· LMArena data1,264.3
254GPT OSS 20b· LMArena data1,258.8
255Llama 4 Maverick· LMArena data1,258.4
256Mistral Small 3.1 24B Instruct2503· LMArena data1,256.8
257Grok 22024-08-13· LMArena data1,253.5
258Granite 4.0 H Small· LMArena data1,252.3
259Hunyuan Large Vision· LMArena data1,251.7
260GLM 4 Plus· LMArena data1,251.6
261GPT 4o2024-05-13· LMArena data1,249.8
262Qwen Max0919· LMArena data1,247.8
263Gemma 3 12B IT· LMArena data1,247.5
264Hunyuan Standard2025-02-10· LMArena data1,247.5
265Claude 3.5 Sonnet20240620· LMArena data1,246.7
266Gemma 3n E4B IT· LMArena data1,245.6
267Qwen2.5 72B Instruct· LMArena data1,245.1
268Gemini 1.5 Pro001· LMArena data1,244.7
269Llama 3.1 405B InstructFP8· LMArena data1,242.8
270Llama 3.1 Nemotron 70B Instruct· LMArena data1,242.1
271GPT 4o2024-08-06· LMArena data1,240.4
272DeepSeek V2.5· LMArena data1,239.5
273Grok 2 mini2024-08-13· LMArena data1,237.6
274Llama 4 Scout· LMArena data1,235.2
275GPT-4o mini2024-07-18· LMArena data1,234.3
276Gemini 1.5 Flash002· LMArena data1,232.8
277Mistral Large 22407· LMArena data1,231.9
278Llama 3.1 405B InstructBF16· LMArena data1,228.8
279Athene 70B0725· LMArena data1,227.5
280Llama 3.3 70B Instruct· LMArena data1,225.2
281Magistral Medium2506· LMArena data1,223.8
282GPT-4 Turbo2024-04-09· LMArena data1,223.1
283Gemma 3 4B IT· LMArena data1,222.8
284Claude 3 Opus20240229· LMArena data1,222.4
285Qwen 2.5 Coder 32B Instruct· LMArena data1,220.3
286Jamba 1.5 Large· LMArena data1,217.4
287Gemini Advanced0514· LMArena data1,214.8
288Reka Core20240904· LMArena data1,213.3
289GPT-4 Turbo Preview1106· LMArena data1,211.1
290Nova Pro 1.0· LMArena data1,210.4
291Llama 3.1 70B Instruct· LMArena data1,208.5
292Mistral Large2411· LMArena data1,207.7
293Claude Haiku 3.520241022· LMArena data1,207.2
294GPT-4 Turbo Preview0125· LMArena data1,205.1
295Phi 4· LMArena data1,202.6
296Mistral Small 24B Instruct2501· LMArena data1,201.6
297Nova Lite 1.0· LMArena data1,201.2
298Gemini 1.5 Flash001· LMArena data1,199.1
299Gemini 1.5 Flash-8B001· LMArena data1,184.9
300Nova Micro 1.0· LMArena data1,183.3

Showing 251–300 of 371 results.

Data credits: LMArena, CC BY 4.0 (licence)