Model leaderboard

Measured on BenchClub test sets

5 models · 4 test sets · 14 runs

<50%50–75%75–90%≥90%* Some answers cut off or failed

Arena scores (LMArena)

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

#ModelScore
1GLM-5.3-FlashZ.ai1503
2GLM-5.3Z.ai1500
3Kimi-K3Moonshot AI1496
4GLM-5.2Z.ai1483
5InklingThinking Machines1481
6MiMo-V2.5-ProXiaomi MiMo1481
7DeepSeek-V4.1-FlashDeepSeek1479
8Hy3Tencent1477
9Kimi-K2.6Moonshot AI1475
10GLM-5.1Z.ai1472
11gemma-4-26B-A4B-itGoogle1470
12Kimi-K2.5Moonshot AI1469
13gemma-4-31B-itGoogle1462
14Beam-1.0-501B-A23Breflection1459
15Inkling-SmallThinking Machines1459
16NVIDIA-Nemotron-3-Ultra-550B-A55B-BF16NVIDIA1456
17Mistral-Large-4-1T-A52BMistral AI1456
18Qwen3.8-27BQwen1455
19Qwen3.5-397B-A17BQwen1453
20DeepSeek-V4-ProDeepSeek1442
21LongCat-Flash-ChatMeituan LongCat1441
22Qwen3-Next-80B-A3B-InstructQwen1440
23GLM-5Z.ai1439
24MiMo-V2.5Xiaomi MiMo1436
25DeepSeek-V3.2DeepSeek1435
26GLM-4.6Z.ai1432
27Qwen3.5-122B-A10BQwen1431
28Qwen3-235B-A22B-Instruct-2507Qwen1431
29Mistral-Medium-3.5-128BMistral AI1430
30Qwen3.5-27BQwen1429
31GLM-4.5Z.ai1428
32Qwen3-VL-235B-A22B-InstructQwen1427
33DeepSeek-V4-FlashDeepSeek1426
34Hy3-previewTencent1424
35MiniMax-M3MiniMax1423
36GLM-4.7Z.ai1423
37MiniMax-M2.7MiniMax1422
38DeepSeek-V3.2-ExpDeepSeek1420
39DeepSeek-V3.1DeepSeek1420
40Mistral-Large-3-675B-Instruct-2512Mistral AI1416
41Qwen3-235B-A22B-Thinking-2507Qwen1414
42NVIDIA-Nemotron-3.5-Super-VL-120B-A12B-BF16NVIDIA1413
43Muse-Glimmer-30Bmeta-models1412
44Qwen3-VL-235B-A22B-ThinkingQwen1410
45Step-3.5-FlashStepFun1406
46Qwen3.5-35B-A3BQwen1403
47DeepSeek-V3.1-TerminusDeepSeek1402
48DeepSeek-R1-0528DeepSeek1400
49Qwen3-32BQwen1399
50Qwen3-Next-80B-A3B-ThinkingQwen1398
51Qwen3-235B-A22BQwen1397
52GLM-4.5-AirZ.ai1396
53MiMo-V2-FlashXiaomi MiMo1395
54Qwen3-30B-A3B-Instruct-2507Qwen1395
55DeepSeek-R1DeepSeek1393
56gpt-oss-120bOpenAI1389
57NVIDIA-Nemotron-3-Super-120B-A12B-BF16NVIDIA1379
58MiniMax-M2.5MiniMax1377
59INTELLECT-3Prime Intellect1376
60DeepSeek-V3-0324DeepSeek1372
61NVIDIA-Nemotron-3-Nano-30B-A3B-BF16NVIDIA1370
62NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16NVIDIA1366
63Qwen3-Coder-480B-A35B-InstructQwen1365
64Trinity-Large-ThinkingArcee AI1365
65Ling-flash-2.0inclusionAI1365
66QwQ-32BQwen1359
67GLM-4.7-FlashZ.ai1355
68Qwen3-30B-A3BQwen1355
69GLM-4.5VZ.ai1354
70MiniMax-M2MiniMax1353
71Ring-flash-2.0inclusionAI1347
72Mistral-Small-3.2-24B-Instruct-2506Mistral AI1342
73Trinity-Large-PreviewArcee AI1336
74gpt-oss-20bOpenAI1317
75Olmo-3-32B-ThinkAi21315
76granite-4.1-8bIBM Granite1313
77gemma-3-27b-itGoogle1312
78DeepSeek-V3DeepSeek1311
79gemma-3-12b-itGoogle1307
80Olmo-3.1-32B-InstructAi21307
81Olmo-3.1-32B-ThinkAi21305
82Llama-4-Maverick-17B-128E-InstructMeta Llama1300
83DeepSeek-V2.5-1210DeepSeek1288
84Llama-4-Scout-17B-16E-InstructMeta Llama1287
85Qwen2.5-72B-InstructQwen1283
86Llama-3.1-405B-InstructMeta Llama1281
87Llama-3.1-Nemotron-70B-InstructNVIDIA1271
88DeepSeek-V2.5DeepSeek1271
89Llama-3.3-70B-InstructMeta Llama1267
90Mistral-Small-3.1-24B-Instruct-2503Mistral AI1261
91granite-4.0-h-smallIBM Granite1253
92Llama-3.1-70B-InstructMeta Llama1252
93Qwen2.5-Coder-32B-InstructQwen1251
94gemma-3n-E4B-itGoogle1250
95phi-4Microsoft1246
96Llama-3.1-Tulu-3-70BAi21242
97DeepSeek-Coder-V2-InstructDeepSeek1241
98Mistral-Small-24B-Instruct-2501Mistral AI1240
99gemma-3-4b-itGoogle1239
100Qwen2-72B-InstructQwen1235
101Meta-Llama-3-70B-InstructMeta Llama1218
102Nemotron-4-340B-InstructNVIDIA1216
103QwQ-32B-PreviewQwen1213
104gemma-2-27b-itGoogle1212
105aya-expanse-32bCohere Labs1200
106Llama-3.1-Tulu-3-8BAi21195
107Qwen1.5-110B-ChatQwen1185
108Mixtral-8x22B-Instruct-v0.1Mistral AI1184
109gemma-2-9b-itGoogle1183
110Llama-3.1-8B-InstructMeta Llama1179
111Phi-3-medium-4k-instructMicrosoft1173
112aya-expanse-8bCohere Labs1168
113Qwen1.5-72B-ChatQwen1164
114granite-3.1-2b-instructIBM Granite1159
115Qwen1.5-32B-ChatQwen1155
116Phi-3-mini-4k-instructMicrosoft1152
117granite-3.1-8b-instructIBM Granite1152
118Meta-Llama-3-8B-InstructMeta Llama1151
119Phi-3-small-8k-instructMicrosoft1151
120Mixtral-8x7B-Instruct-v0.1Mistral AI1147
121granite-3.0-8b-instructIBM Granite1143
122gemma-2-2b-itGoogle1135
123Llama-3.2-3B-InstructMeta Llama1126
124Qwen1.5-14B-ChatQwen1125
125granite-3.0-2b-instructIBM Granite1117
126deepseek-llm-67b-chatDeepSeek1108
127gemma-1.1-7b-itGoogle1107
128Llama-2-70b-chat-hfMeta Llama1091
129Phi-3-mini-128k-instructMicrosoft1089
130Llama-3.2-1B-InstructMeta Llama1086
131Mistral-7B-Instruct-v0.2Mistral AI1085
132Qwen1.5-7B-ChatQwen1080
133Qwen-14B-ChatQwen1068
134gemma-7b-itGoogle1066
135Llama-2-13b-chat-hfMeta Llama1065
136SOLAR-10.7B-Instruct-v1.0Upstage1064
137gemma-1.1-2b-itGoogle1047
138Llama-2-7b-chat-hfMeta Llama1042
139Qwen1.5-4B-ChatQwen1026
140OLMo-7B-InstructAi21018
141gemma-2b-itGoogle1009
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.