Skip to content

Benchmark

Arena ELO - Industry Entertainment and Sports and Media

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

Models measured
364
Scores recorded
421
Sources
1
IQ weight
1.17

Score distribution

1,279.8
Mean
1,314.7
Median
149.5
Std dev
1,058.5
10th percentile
1,441.1
90th percentile
784.8
Min
784.81,526.7
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.17 for chat models, measured over 364 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard

#ModelScore
301GPT OSS 20b· LMArena data1,216.7
302Granite 4.2 3B· LMArena data1,215.1
303Mercury· LMArena data1,214.8
304Gemma 2 9B IT SimPO· LMArena data1,214.7
305Granite 4.0 H Small· LMArena data1,213
306Gemma 2 27B IT· LMArena data1,212.5
307Qwen2.5 72B Instruct· LMArena data1,211
308Nova Pro 1.0· LMArena data1,209.8
309Nemotron-4 340B Instruct· LMArena data1,207.8
310Gemini 1.5 Flash001· LMArena data1,207.6
311Reka Core20240904· LMArena data1,207.3
312Llama 3 70B Instruct· LMArena data1,195.4
313Llama 3.1 Nemotron 51B Instruct· LMArena data1,191
314Command R+· LMArena data1,189.6
315GLM 40520· LMArena data1,185.3
316Gemini 1.5 Flash-8B001· LMArena data1,185.1
317Gemma 2 9B IT· LMArena data1,184.9
318GPT 40314· LMArena data1,184.4
319Mistral Small 24B Instruct2501· LMArena data1,183.9
320Aya Expanse 32B· LMArena data1,183.5
321GPT 40613· LMArena data1,180
322Reka Flash20240904· LMArena data1,177.2
323Nova Lite 1.0· LMArena data1,175.7
324Claude 3 Sonnet20240229· LMArena data1,171.4
325Qwen2 72B Instruct· LMArena data1,165.6
326Nova Micro 1.0· LMArena data1,163
327Phi 4· LMArena data1,162
328Llama 3.1 Tulu 3 8B· LMArena data1,160.9
329Jamba 1.5 Mini· LMArena data1,158.2
330Hunyuan Standard 256K· LMArena data1,152.9
331Claude 3 Haiku20240307· LMArena data1,151.9
332Mistral Large2402· LMArena data1,150.8
333Qwen 2.5 Coder 32B Instruct· LMArena data1,150.2
334Aya Expanse 8B· LMArena data1,149.9
335Llama 3.1 8B Instruct· LMArena data1,144.7
336Ministral 8B2410· LMArena data1,140.1
337Mixtral 8x22B Instruct v0.1· LMArena data1,138.7
338Qwen1.5-110B-Chat· LMArena data1,138.6
339Mistral Medium· LMArena data1,137.8
340Reka Flash 21BOnline, 20240226· LMArena data1,133.9
341Llama 3 8B Instruct· LMArena data1,132
342Zephyr ORPO 141B A35B v0.1· LMArena data1,128.2
343Gemma 2 2B IT· LMArena data1,126.9
344Qwen1.5-72B-Chat· LMArena data1,125.5
345DeepSeek Coder V2 Instruct· LMArena data1,122.2
346Yi 1.5 34B Chat· LMArena data1,120.7
347Tulu 2 DPO 70B· LMArena data1,115.6
348QwQ 32B Preview· LMArena data1,114.6
349WizardLM 70B· LMArena data1,112.3
350Granite 3.1 8B Instruct· LMArena data1,111.2

Showing 301–350 of 421 results.

Data credits: LMArena, CC BY 4.0 (licence)