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

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

#ModelScore
1GLM-5.3-FlashZ.ai1503
2GLM-5.3Z.ai1500
3GLM-5.2Z.ai1483
4InklingThinking Machines1481
5MiMo-V2.5-ProXiaomi MiMo1481
6DeepSeek-V4.1-FlashDeepSeek1479
7Hy3Tencent1477
8Kimi-K2.6Moonshot AI1475
9GLM-5.1Z.ai1472
10gemma-4-26B-A4B-itGoogle1470
11Kimi-K2.5Moonshot AI1469
12gemma-4-31B-itGoogle1462
13Beam-1.0-501B-A23Breflection1459
14Inkling-SmallThinking Machines1459
15NVIDIA-Nemotron-3-Ultra-550B-A55B-BF16NVIDIA1456
16Mistral-Large-4-1T-A52BMistral AI1456
17Qwen3.8-27BQwen1455
18Qwen3.5-397B-A17BQwen1453
19LongCat-Flash-ChatMeituan LongCat1441
20Qwen3-Next-80B-A3B-InstructQwen1440
21GLM-5Z.ai1439
22MiMo-V2.5Xiaomi MiMo1436
23GLM-4.6Z.ai1432
24Qwen3.5-122B-A10BQwen1431
25Qwen3-235B-A22B-Instruct-2507Qwen1431
26Mistral-Medium-3.5-128BMistral AI1430
27Qwen3.5-27BQwen1429
28GLM-4.5Z.ai1428
29Qwen3-VL-235B-A22B-InstructQwen1427
30Hy3-previewTencent1424
31MiniMax-M3MiniMax1423
32GLM-4.7Z.ai1423
33MiniMax-M2.7MiniMax1422
34Mistral-Large-3-675B-Instruct-2512Mistral AI1416
35Qwen3-235B-A22B-Thinking-2507Qwen1414
36NVIDIA-Nemotron-3.5-Super-VL-120B-A12B-BF16NVIDIA1413
37Muse-Glimmer-30Bmeta-models1412
38Qwen3-VL-235B-A22B-ThinkingQwen1410
39Step-3.5-FlashStepFun1406
40Qwen3.5-35B-A3BQwen1403
41Qwen3-32BQwen1399
42Qwen3-Next-80B-A3B-ThinkingQwen1398
43Qwen3-235B-A22BQwen1397
44GLM-4.5-AirZ.ai1396
45MiMo-V2-FlashXiaomi MiMo1395
46Qwen3-30B-A3B-Instruct-2507Qwen1395
47gpt-oss-120bOpenAI1389
48NVIDIA-Nemotron-3-Super-120B-A12B-BF16NVIDIA1379
49MiniMax-M2.5MiniMax1377
50INTELLECT-3Prime Intellect1376
51NVIDIA-Nemotron-3-Nano-30B-A3B-BF16NVIDIA1370
52NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16NVIDIA1366
53Qwen3-Coder-480B-A35B-InstructQwen1365
54Trinity-Large-ThinkingArcee AI1365
55Ling-flash-2.0inclusionAI1365
56QwQ-32BQwen1359
57GLM-4.7-FlashZ.ai1355
58Qwen3-30B-A3BQwen1355
59GLM-4.5VZ.ai1354
60MiniMax-M2MiniMax1353
61Trinity-Large-PreviewArcee AI1336
62gpt-oss-20bOpenAI1317
63granite-4.1-8bIBM Granite1313
64Olmo-3.1-32B-InstructAi21307
65Llama-4-Maverick-17B-128E-InstructMeta Llama1300
66Llama-4-Scout-17B-16E-InstructMeta Llama1287
67Qwen2.5-72B-InstructQwen1283
68Llama-3.1-405B-InstructMeta Llama1281
69Llama-3.3-70B-InstructMeta Llama1267
70granite-4.0-h-smallIBM Granite1253
71Llama-3.1-70B-InstructMeta Llama1252
72Qwen2.5-Coder-32B-InstructQwen1251
73QwQ-32B-PreviewQwen1213
74Mixtral-8x22B-Instruct-v0.1Mistral AI1184
75Llama-3.1-8B-InstructMeta Llama1179
76granite-3.1-2b-instructIBM Granite1159
77granite-3.1-8b-instructIBM Granite1152
78granite-3.0-8b-instructIBM Granite1143
79Llama-3.2-3B-InstructMeta Llama1126
80granite-3.0-2b-instructIBM Granite1117
81Llama-3.2-1B-InstructMeta Llama1086
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.