Skip to content

Benchmark

Arena ELO - Hard Prompts

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

Models measured
366
Scores recorded
423
Sources
1
IQ weight
1.24

Score distribution

1,308.5
Mean
1,356.7
Median
162.3
Std dev
1,065.4
10th percentile
1,480.3
90th percentile
727.9
Min
727.91,550.7
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.24 for chat models, measured over 366 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 3.5 Sonnet20241022· LMArena data1,305
252Granite 4.2 3B· LMArena data1,304.5
253Yi Lightning· LMArena data1,302
254Olmo 3 32B Think· LMArena data1,301.9
255Qwen2.5-Plus1127· LMArena data1,299.2
256Step-2 16K Exp202412· LMArena data1,299
257Gemini 1.5 Pro002· LMArena data1,296.2
258GLM 4 Plus0111· LMArena data1,293.5
259Granite 4.1 8B· LMArena data1,293.3
260Athene V2 Chat· LMArena data1,292.4
261DeepSeek V2.51210· LMArena data1,289.1
262Molmo 2 8B· LMArena data1,287.5
263Hunyuan Large2025-02-10· LMArena data1,286.2
264GPT-4.1 nano2025-04-14· LMArena data1,285.5
265Mercury· LMArena data1,284.4
266Gemma 3n E4B IT· LMArena data1,283.8
267GPT 4o2024-05-13· LMArena data1,280.7
268Llama 4 Maverick· LMArena data1,280.6
269Mistral Small 3.1 24B Instruct2503· LMArena data1,277.7
270Claude 3.5 Sonnet20240620· LMArena data1,275.3
271GPT OSS 20b· LMArena data1,273.9
272GLM 4 Plus· LMArena data1,272.9
273Grok 22024-08-13· LMArena data1,272
274Olmo 3.1 32B Think· LMArena data1,271.7
275Qwen2.5 72B Instruct· LMArena data1,270.5
276DeepSeek V2.5· LMArena data1,270.2
277Qwen Max0919· LMArena data1,269.3
278Llama 3.1 405B InstructBF16· LMArena data1,269.3
279GPT-4o mini2024-07-18· LMArena data1,266.9
280Magistral Medium2506· LMArena data1,266.7
281Llama 3.1 Nemotron 70B Instruct· LMArena data1,266.2
282Llama 4 Scout· LMArena data1,265.9
283Hunyuan Standard2025-02-10· LMArena data1,264.2
284GPT 4o2024-08-06· LMArena data1,263.9
285Llama 3.1 405B InstructFP8· LMArena data1,263.4
286Gemini 1.5 Flash002· LMArena data1,257.1
287Llama 3.3 70B Instruct· LMArena data1,257.1
288Mistral Large2411· LMArena data1,256.7
289Hunyuan Large Vision· LMArena data1,256.5
290Mistral Large 22407· LMArena data1,256.4
291Grok 2 mini2024-08-13· LMArena data1,254.8
292Gemini 1.5 Pro001· LMArena data1,254.8
293Gemma 3 4B IT· LMArena data1,252.5
294Qwen 2.5 Coder 32B Instruct· LMArena data1,251.2
295GPT-4 Turbo2024-04-09· LMArena data1,250.8
296Claude Haiku 3.520241022· LMArena data1,250.8
297Gemini Advanced0514· LMArena data1,248.1
298Nova Pro 1.0· LMArena data1,245.4
299Claude 3 Opus20240229· LMArena data1,245.3
300Athene 70B0725· LMArena data1,243

Showing 251–300 of 423 results.

Data credits: LMArena, CC BY 4.0 (licence)