Skip to content

Benchmark

Arena ELO - Multi Turn

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

Models measured
364
Scores recorded
421
Sources
1
IQ weight
1.21

Score distribution

1,304.7
Mean
1,350
Median
160.8
Std dev
1,057.9
10th percentile
1,467.2
90th percentile
740.3
Min
740.31,569.6
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.21 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-4 Turbo Preview0125· LMArena data1,250.4
302Nova Pro 1.0· LMArena data1,245.5
303Granite 4.0 H Small· LMArena data1,243.3
304Gemini 1.5 Flash001· LMArena data1,227.5
305Llama 3.1 Nemotron 51B Instruct· LMArena data1,226.8
306Claude 3 Sonnet20240229· LMArena data1,226.8
307Reka Core20240904· LMArena data1,226.5
308Gemma 2 27B IT· LMArena data1,223.7
309Llama 3 70B Instruct· LMArena data1,222.9
310Qwen 2.5 Coder 32B Instruct· LMArena data1,222.3
311OLMo 2 32B Instruct0325· LMArena data1,221.2
312Jamba 1.5 Large· LMArena data1,219.8
313Gemma 2 9B IT SimPO· LMArena data1,219.6
314GLM 40520· LMArena data1,217.9
315Mistral Small 24B Instruct2501· LMArena data1,216.6
316Nemotron-4 340B Instruct· LMArena data1,213.7
317Command R+08-2024· LMArena data1,212.7
318GPT 40314· LMArena data1,205.9
319Phi 4· LMArena data1,205.9
320Nova Lite 1.0· LMArena data1,198.6
321Qwen2 72B Instruct· LMArena data1,195.5
322Gemma 2 9B IT· LMArena data1,193.4
323Command R+· LMArena data1,192.1
324Aya Expanse 32B· LMArena data1,190
325Claude 3 Haiku20240307· LMArena data1,189.9
326Reka Flash20240904· LMArena data1,188.8
327Gemini 1.5 Flash-8B001· LMArena data1,185.2
328GPT 40613· LMArena data1,184.7
329Nova Micro 1.0· LMArena data1,177.9
330DeepSeek Coder V2 Instruct· LMArena data1,176.5
331Hunyuan Standard 256K· LMArena data1,174.6
332Mistral Large2402· LMArena data1,172
333Llama 3.1 8B Instruct· LMArena data1,172
334Ministral 8B2410· LMArena data1,166.3
335Qwen1.5-72B-Chat· LMArena data1,159.9
336Qwen1.5-110B-Chat· LMArena data1,159.7
337Aya Expanse 8B· LMArena data1,159.6
338Jamba 1.5 Mini· LMArena data1,156.9
339Llama 3.1 Tulu 3 8B· LMArena data1,153.6
340Yi 1.5 34B Chat· LMArena data1,153
341Llama 3 8B Instruct· LMArena data1,152.4
342Reka Flash 21BOnline, 20240226· LMArena data1,151
343Mistral Medium· LMArena data1,147.2
344Reka Flash 21B20240226· LMArena data1,141.5
345QwQ 32B Preview· LMArena data1,140.6
346Qwen1.5-32B-Chat· LMArena data1,139.7
347Gemini Pro· LMArena data1,139.4
348InternLM2.5 20B Chat· LMArena data1,135.5
349Mixtral 8x22B Instruct v0.1· LMArena data1,130.1
350Gemini ProDev API· LMArena data1,124.2

Showing 301–350 of 421 results.

Data credits: LMArena, CC BY 4.0 (licence)