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

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

#ModelScore
1GLM-5.3Z.ai1519
2GLM-5.3-FlashZ.ai1511
3DeepSeek-V4.1-FlashDeepSeek1503
4MiMo-V2.5-ProXiaomi MiMo1502
5Kimi-K2.6Moonshot AI1492
6GLM-5.2Z.ai1488
7Qwen3.8-27BQwen1479
8Inkling-2-SmallThinking Machines1478
9GLM-5.1Z.ai1477
10NVIDIA-Nemotron-3-Ultra-550B-A55B-BF16NVIDIA1473
11InklingThinking Machines1469
12Kimi-K2.5Moonshot AI1465
13gemma-4-31B-itGoogle1465
14Beam-1.0-501B-A23Breflection1464
15Qwen3-235B-A22B-Thinking-2507Qwen1463
16Qwen3.5-397B-A17BQwen1462
17MiniMax-M3MiniMax1460
18MiMo-V2.5Xiaomi MiMo1460
19Hy3Tencent1458
20GLM-5Z.ai1455
21gemma-4-26B-A4B-itGoogle1448
22Mistral-Large-4-1T-A52BMistral AI1446
23MiniMax-M2.7MiniMax1443
24Qwen3-VL-235B-A22B-InstructQwen1442
25Qwen3-235B-A22B-Instruct-2507Qwen1442
26LongCat-Flash-ChatMeituan LongCat1441
27Inkling-SmallThinking Machines1440
28GLM-4.5Z.ai1433
29GLM-4.6Z.ai1433
30Mistral-Medium-3.5-128BMistral AI1431
31Qwen3.5-122B-A10BQwen1431
32Qwen3-VL-235B-A22B-ThinkingQwen1429
33Qwen3.5-27BQwen1428
34MiMo-V2-FlashXiaomi MiMo1426
35Hy3-previewTencent1424
36GLM-4.7Z.ai1424
37Mistral-Large-3-675B-Instruct-2512Mistral AI1422
38NVIDIA-Nemotron-3.5-Super-VL-120B-A12B-BF16NVIDIA1422
39Step-3.5-FlashStepFun1421
40Qwen3-Next-80B-A3B-InstructQwen1417
41granite-4.2-30bIBM Granite1408
42Qwen3.5-35B-A3BQwen1408
43NVIDIA-Nemotron-3-Super-120B-A12B-BF16NVIDIA1399
44Qwen3-30B-A3B-Instruct-2507Qwen1396
45Muse-Glimmer-30Bmeta-models1388
46granite-4.2-8bIBM Granite1381
47MiniMax-M2.5MiniMax1380
48Qwen3-Next-80B-A3B-ThinkingQwen1375
49GLM-4.6VZ.ai1372
50GLM-4.5-AirZ.ai1370
51Qwen3-235B-A22BQwen1369
52Qwen3-32BQwen1362
53NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16NVIDIA1360
54Trinity-Large-ThinkingArcee AI1359
55GLM-4.7-FlashZ.ai1358
56gpt-oss-120bOpenAI1355
57NVIDIA-Nemotron-3-Nano-30B-A3B-BF16NVIDIA1355
58Ling-flash-2.0inclusionAI1353
59Trinity-Large-PreviewArcee AI1352
60GLM-4.5VZ.ai1351
61MiniMax-M2MiniMax1340
62Qwen3-Coder-480B-A35B-InstructQwen1338
63QwQ-32BQwen1323
64INTELLECT-3Prime Intellect1323
65Qwen3-30B-A3BQwen1314
66granite-4.1-8bIBM Granite1312
67Olmo-3.1-32B-InstructAi21306
68granite-4.2-3bIBM Granite1305
69gpt-oss-20bOpenAI1259
70Llama-4-Maverick-17B-128E-InstructMeta Llama1258
71granite-4.0-h-smallIBM Granite1251
72Qwen2.5-72B-InstructQwen1245
73Llama-3.1-405B-InstructMeta Llama1243
74Llama-4-Scout-17B-16E-InstructMeta Llama1235
75Llama-3.3-70B-InstructMeta Llama1225
76Qwen2.5-Coder-32B-InstructQwen1220
77Llama-3.1-70B-InstructMeta Llama1208
78Llama-3.1-8B-InstructMeta Llama1144
79granite-3.1-8b-instructIBM Granite1141
80QwQ-32B-PreviewQwen1134
81granite-3.1-2b-instructIBM Granite1131
82Mixtral-8x22B-Instruct-v0.1Mistral AI1113
83Llama-3.2-3B-InstructMeta Llama1090
84granite-3.0-8b-instructIBM Granite1087
85granite-3.0-2b-instructIBM Granite1064
86Llama-3.2-1B-InstructMeta Llama1007
Show the top 20

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.