Skip to content

Benchmark

Arena ELO - Hard Prompts English

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

Models measured
364
Scores recorded
421
Sources
1
IQ weight
1.31

Score distribution

1,319.1
Mean
1,374
Median
157.7
Std dev
1,085.3
10th percentile
1,479.2
90th percentile
734.7
Min
734.71,547.3
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.31 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
251DeepSeek V3· LMArena data1,323.3
252Yi Lightning· LMArena data1,322.1
253Claude Sonnet 3.720250219· LMArena data1,320.1
254Gemma 3 12B IT· LMArena data1,318.1
255Qwen2.5-Plus1127· LMArena data1,317.6
256Granite 4.1 8B· LMArena data1,313.2
257Claude 3.5 Sonnet20241022· LMArena data1,311.4
258Step-2 16K Exp202412· LMArena data1,308.5
259Mercury· LMArena data1,307.9
260Athene V2 Chat· LMArena data1,303.9
261DeepSeek V2.51210· LMArena data1,298.9
262Gemini 1.5 Pro002· LMArena data1,298.7
263Llama 4 Maverick· LMArena data1,295.3
264Hunyuan Large2025-02-10· LMArena data1,294.4
265Llama 4 Scout· LMArena data1,293
266GPT 4o2024-05-13· LMArena data1,292.7
267GPT-4.1 nano2025-04-14· LMArena data1,292.5
268GPT OSS 20b· LMArena data1,292.1
269GLM 4 Plus0111· LMArena data1,291.5
270Gemma 3n E4B IT· LMArena data1,291.4
271Llama 3.1 405B InstructBF16· LMArena data1,291.4
272Mistral Small 3.1 24B Instruct2503· LMArena data1,289.4
273Magistral Medium2506· LMArena data1,284.4
274Llama 3.1 405B InstructFP8· LMArena data1,283.2
275Llama 3.1 Nemotron 70B Instruct· LMArena data1,282.6
276Claude 3.5 Sonnet20240620· LMArena data1,282.4
277Qwen Max0919· LMArena data1,280.6
278Grok 22024-08-13· LMArena data1,279.7
279DeepSeek V2.5· LMArena data1,279.6
280GLM 4 Plus· LMArena data1,279.6
281Qwen2.5 72B Instruct· LMArena data1,279.4
282GPT-4o mini2024-07-18· LMArena data1,278.4
283GPT 4o2024-08-06· LMArena data1,276.9
284Llama 3.3 70B Instruct· LMArena data1,276.7
285Hunyuan Standard2025-02-10· LMArena data1,273.9
286Hunyuan Large Vision· LMArena data1,270.8
287Qwen 2.5 Coder 32B Instruct· LMArena data1,269.2
288Mistral Large 22407· LMArena data1,269.1
289Mistral Large2411· LMArena data1,267.7
290Grok 2 mini2024-08-13· LMArena data1,267.5
291Granite 4.0 H Small· LMArena data1,267.1
292GPT-4 Turbo2024-04-09· LMArena data1,266
293Gemini 1.5 Flash002· LMArena data1,264.7
294Claude Haiku 3.520241022· LMArena data1,264.2
295Llama 3.1 70B Instruct· LMArena data1,261.2
296Gemini 1.5 Pro001· LMArena data1,258.8
297Nova Pro 1.0· LMArena data1,258.7
298Athene 70B0725· LMArena data1,254.7
299GPT-4 Turbo Preview1106· LMArena data1,253.8
300Gemma 3 4B IT· LMArena data1,253.2

Showing 251–300 of 421 results.

Data credits: LMArena, CC BY 4.0 (licence)