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

32 models in hard prompts ranked by LMArena score · updated Nov 11, 2026

#ModelScore
32gemma-3-4b-itGoogle1253
31Llama-4-Scout-17B-16E-InstructMeta Llama1266
30Llama-4-Maverick-17B-128E-InstructMeta Llama1281
29gemma-3n-E4B-itGoogle1284
28Molmo2-8BAi21287
27gemma-3-12b-itGoogle1309
26GLM-4.5VZ.ai1334
25gemma-3-27b-itGoogle1340
24GLM-4.6VZ.ai1368
23Muse-Glimmer-30Bmeta-models1395
22Qwen3.5-35B-A3BQwen1400
21Qwen3-VL-235B-A22B-ThinkingQwen1407
20NVIDIA-Nemotron-3.5-Super-VL-120B-A12B-BF16NVIDIA1414
19Qwen3.5-27BQwen1414
18Inkling-2-SmallThinking Machines1421
17Qwen3.5-122B-A10BQwen1421
16Inkling-SmallThinking Machines1425
15Qwen3-VL-235B-A22B-InstructQwen1428
14Mistral-Large-4-1T-A52BMistral AI1434
13gemma-4-26B-A4B-itGoogle1439
Show all 32 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.