Skip to content

Benchmark

Arena ELO - Math

Math benchmarks measure how well models solve maths problems. Higher scores are better.

Models measured
350
Scores recorded
405
Sources
1
IQ weight
1.24

Score distribution

1,316
Mean
1,367.4
Median
152.6
Std dev
1,089
10th percentile
1,474.8
90th percentile
837.8
Min
837.81,547.6
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 350 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard

#ModelScore
251GLM 4 Plus0111· LMArena data1,287.4
252Llama 4 Scout· LMArena data1,287
253GPT 4o2024-08-06· LMArena data1,285.4
254GPT 4o2024-05-13· LMArena data1,284.4
255Grok 22024-08-13· LMArena data1,283.2
256Qwen2.5 72B Instruct· LMArena data1,282.6
257Llama 3.1 405B InstructFP8· LMArena data1,281.3
258Hunyuan Large2025-02-10· LMArena data1,280.8
259Llama 3.1 405B InstructBF16· LMArena data1,278.3
260Qwen Max0919· LMArena data1,275
261GLM 4 Plus· LMArena data1,274.9
262GPT-4.1 nano2025-04-14· LMArena data1,274.3
263Hunyuan Standard2025-02-10· LMArena data1,274.2
264Hunyuan Turbo0110· LMArena data1,272.9
265Claude 3 Opus20240229· LMArena data1,272.6
266Gemini Advanced0514· LMArena data1,271.7
267GPT-4 Turbo2024-04-09· LMArena data1,271.6
268Llama 3.1 Nemotron 70B Instruct· LMArena data1,271
269DeepSeek V2.5· LMArena data1,270.7
270Gemini 1.5 Pro001· LMArena data1,269.5
271GPT-4 Turbo Preview1106· LMArena data1,269.2
272Gemini 1.5 Flash002· LMArena data1,269.1
273Hunyuan Large Vision· LMArena data1,268.1
274GPT-4 Turbo Preview0125· LMArena data1,267.8
275GPT-4o mini2024-07-18· LMArena data1,267.4
276Llama 3.3 70B Instruct· LMArena data1,267.2
277Grok 2 mini2024-08-13· LMArena data1,265.2
278Mistral Large 22407· LMArena data1,261.5
279Mistral Large2411· LMArena data1,261.1
280Mistral Small 3.1 24B Instruct2503· LMArena data1,261
281Granite 4.0 H Small· LMArena data1,255.3
282Llama 3.1 70B Instruct· LMArena data1,251.7
283Nova Pro 1.0· LMArena data1,251.6
284Magistral Medium2506· LMArena data1,251
285Gemma 3n E4B IT· LMArena data1,250.8
286Qwen 2.5 Coder 32B Instruct· LMArena data1,250.5
287Phi 4· LMArena data1,245.8
288Claude Haiku 3.520241022· LMArena data1,243.8
289Llama 3.1 Tulu 3 70B· LMArena data1,241.9
290DeepSeek Coder V2 Instruct· LMArena data1,241.3
291Mistral Small 24B Instruct2501· LMArena data1,240.2
292Gemma 3 4B IT· LMArena data1,239.1
293Qwen2 72B Instruct· LMArena data1,234.8
294Hunyuan Standard 256K· LMArena data1,234.7
295Athene 70B0725· LMArena data1,231.4
296GPT 40314· LMArena data1,230.1
297Llama 3.1 Nemotron 51B Instruct· LMArena data1,229.9
298Gemini 1.5 Flash001· LMArena data1,228.7
299Nova Lite 1.0· LMArena data1,226.6
300Reka Core20240904· LMArena data1,221.7

Showing 251–300 of 405 results.

Data credits: LMArena, CC BY 4.0 (licence)