Through an API

Open and closed models answer the same suite-standard 1.0 through their APIs, ranked together. 1 ranked model.

#ModelPass rate
1DeepSeek V4.1 Flash 1017Open weights100%

Model leaderboard

28 models in math ranked by LMArena score · updated Nov 11, 2026

#ModelScore
28gemma-3-4b-itGoogle1239
27gemma-3n-E4B-itGoogle1250
26Llama-4-Scout-17B-16E-InstructMeta Llama1287
25Llama-4-Maverick-17B-128E-InstructMeta Llama1300
24gemma-3-12b-itGoogle1307
23gemma-3-27b-itGoogle1312
22GLM-4.5VZ.ai1354
21Qwen3.5-35B-A3BQwen1403
20Qwen3-VL-235B-A22B-ThinkingQwen1410
19Muse-Glimmer-30Bmeta-models1412
18NVIDIA-Nemotron-3.5-Super-VL-120B-A12B-BF16NVIDIA1413
17MiniMax-M3MiniMax1423
16Qwen3-VL-235B-A22B-InstructQwen1427
15Qwen3.5-27BQwen1429
14Qwen3.5-122B-A10BQwen1431
13MiMo-V2.5Xiaomi MiMo1436
12Qwen3.5-397B-A17BQwen1453
11Qwen3.8-27BQwen1455
10Mistral-Large-4-1T-A52BMistral AI1456
9Inkling-SmallThinking Machines1459
Show all 28 models

Score by size

Plot against
Best for its size

Scores: LMArena, CC BY 4.0. Measured on served versions, not local builds. A model with several served variants shows its best.