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

31 models in expert ranked by LMArena score · updated Nov 11, 2026

#ModelScore
31gemma-3-4b-itGoogle1223
30Llama-4-Scout-17B-16E-InstructMeta Llama1235
29gemma-3n-E4B-itGoogle1246
28gemma-3-12b-itGoogle1247
27Llama-4-Maverick-17B-128E-InstructMeta Llama1258
26gemma-3-27b-itGoogle1303
25GLM-4.5VZ.ai1351
24GLM-4.6VZ.ai1372
23Muse-Glimmer-30Bmeta-models1388
22Qwen3.5-35B-A3BQwen1408
21NVIDIA-Nemotron-3.5-Super-VL-120B-A12B-BF16NVIDIA1422
20Qwen3.5-27BQwen1428
19Qwen3-VL-235B-A22B-ThinkingQwen1429
18Qwen3.5-122B-A10BQwen1431
17Inkling-SmallThinking Machines1440
16Qwen3-VL-235B-A22B-InstructQwen1442
15Mistral-Large-4-1T-A52BMistral AI1446
14gemma-4-26B-A4B-itGoogle1448
13MiMo-V2.5Xiaomi MiMo1460
12MiniMax-M3MiniMax1460
Show all 31 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.