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)

148 models in instructions ranked by LMArena score · updated Nov 11, 2026

#ModelScore
1Kimi-K3Moonshot AI1484
2GLM-5.3Z.ai1483
3GLM-5.3-FlashZ.ai1480
4MiMo-V2.5-ProXiaomi MiMo1476
5DeepSeek-V4.1-FlashDeepSeek1473
6Kimi-K3.1-ProMoonshot AI1472
7GLM-5.2Z.ai1467
8GLM-5.1Z.ai1452
9Kimi-K2.6Moonshot AI1451
10DeepSeek-V4-ProDeepSeek1448
11MiMo-V2.5Xiaomi MiMo1434
12gemma-4-31B-itGoogle1433
13Qwen3.8-27BQwen1431
14Kimi-K2.5Moonshot AI1431
15GLM-5Z.ai1428
16MiniMax-M3MiniMax1428
17InklingThinking Machines1427
18Hy3Tencent1426
19Beam-1.0-501B-A23Breflection1424
20Qwen3.5-397B-A17BQwen1424
21DeepSeek-V4-FlashDeepSeek1421
22gemma-4-26B-A4B-itGoogle1421
23NVIDIA-Nemotron-3-Ultra-550B-A55B-BF16NVIDIA1418
24Mistral-Medium-3.5-128BMistral AI1416
25DeepSeek-V3.2DeepSeek1413
26GLM-4.7Z.ai1412
27GLM-4.6Z.ai1410
28Mistral-Large-4-1T-A52BMistral AI1409
29Qwen3-235B-A22B-Instruct-2507Qwen1408
30Inkling-2-SmallThinking Machines1408
31Qwen3-VL-235B-A22B-InstructQwen1406
32MiniMax-M2.7MiniMax1405
33Mistral-Large-3-675B-Instruct-2512Mistral AI1405
34DeepSeek-V3.1-TerminusDeepSeek1404
35GLM-4.5Z.ai1404
36Inkling-SmallThinking Machines1402
37DeepSeek-V3.2-ExpDeepSeek1401
38DeepSeek-V3.1DeepSeek1400
39NVIDIA-Nemotron-3.5-Super-VL-120B-A12B-BF16NVIDIA1400
40Qwen3.5-122B-A10BQwen1399
41LongCat-Flash-ChatMeituan LongCat1397
42MiMo-V2-FlashXiaomi MiMo1393
43Qwen3.5-27BQwen1392
44Qwen3-Next-80B-A3B-InstructQwen1389
45Qwen3-235B-A22B-Thinking-2507Qwen1386
46Step-3.5-FlashStepFun1385
47Hy3-previewTencent1384
48DeepSeek-R1-0528DeepSeek1381
49Qwen3.5-35B-A3BQwen1380
50Qwen3-VL-235B-A22B-ThinkingQwen1374
51Muse-Glimmer-30Bmeta-models1373
52Qwen3-30B-A3B-Instruct-2507Qwen1363
53Qwen3-235B-A22BQwen1362
54DeepSeek-R1DeepSeek1357
55Qwen3-Coder-480B-A35B-InstructQwen1355
56GLM-4.5-AirZ.ai1353
57MiniMax-M2.5MiniMax1353
58GLM-4.6VZ.ai1351
59granite-4.2-30bIBM Granite1350
60NVIDIA-Nemotron-3-Super-120B-A12B-BF16NVIDIA1347
61DeepSeek-V3-0324DeepSeek1345
62Qwen3-Next-80B-A3B-ThinkingQwen1344
63Trinity-Large-PreviewArcee AI1334
64MiniMax-M2MiniMax1329
65Trinity-Large-ThinkingArcee AI1326
66GLM-4.7-FlashZ.ai1326
67gemma-3-27b-itGoogle1321
68NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16NVIDIA1318
69Ling-flash-2.0inclusionAI1318
70gpt-oss-120bOpenAI1318
71INTELLECT-3Prime Intellect1316
72DeepSeek-V3DeepSeek1315
73Ring-flash-2.0inclusionAI1313
74GLM-4.5VZ.ai1310
75Mistral-Small-3.2-24B-Instruct-2506Mistral AI1309
76Qwen3-32BQwen1305
77NVIDIA-Nemotron-3-Nano-30B-A3B-BF16NVIDIA1304
78gemma-3-12b-itGoogle1299
79Olmo-3.1-32B-InstructAi21299
80granite-4.2-8bIBM Granite1299
81QwQ-32BQwen1297
82Qwen3-30B-A3BQwen1283
83DeepSeek-V2.5-1210DeepSeek1279
84Olmo-3-32B-ThinkAi21274
85granite-4.2-3bIBM Granite1272
86Molmo2-8BAi21270
87granite-4.1-8bIBM Granite1269
88Llama-4-Maverick-17B-128E-InstructMeta Llama1267
89Mistral-Small-3.1-24B-Instruct-2503Mistral AI1263
90Llama-3.1-405B-InstructMeta Llama1258
91gemma-3n-E4B-itGoogle1255
92Qwen2.5-72B-InstructQwen1254
93Llama-3.1-Nemotron-70B-InstructNVIDIA1252
94DeepSeek-V2.5DeepSeek1249
95Llama-4-Scout-17B-16E-InstructMeta Llama1248
96Olmo-3.1-32B-ThinkAi21246
97Llama-3.3-70B-InstructMeta Llama1242
98gemma-3-4b-itGoogle1239
99gpt-oss-20bOpenAI1236
100Llama-3.1-Tulu-3-70BAi21232
101Llama-3.1-70B-InstructMeta Llama1231
102Qwen2.5-Coder-32B-InstructQwen1223
103granite-4.0-h-smallIBM Granite1222
104Mistral-Small-24B-Instruct-2501Mistral AI1214
105gemma-2-27b-itGoogle1206
106Nemotron-4-340B-InstructNVIDIA1203
107phi-4Microsoft1201
108aya-expanse-32bCohere Labs1196
109Meta-Llama-3-70B-InstructMeta Llama1194
110Qwen2-72B-InstructQwen1181
111DeepSeek-Coder-V2-InstructDeepSeek1180
112gemma-2-9b-itGoogle1178
113Llama-3.1-Tulu-3-8BAi21174
114Llama-3.1-8B-InstructMeta Llama1159
115Qwen1.5-110B-ChatQwen1158
116QwQ-32B-PreviewQwen1156
117aya-expanse-8bCohere Labs1155
118Mixtral-8x22B-Instruct-v0.1Mistral AI1147
119Qwen1.5-72B-ChatQwen1141
120granite-3.1-8b-instructIBM Granite1131
121Meta-Llama-3-8B-InstructMeta Llama1127
122gemma-2-2b-itGoogle1117
123granite-3.1-2b-instructIBM Granite1116
124Qwen1.5-32B-ChatQwen1115
125Phi-3-medium-4k-instructMicrosoft1113
126Mixtral-8x7B-Instruct-v0.1Mistral AI1109
127Qwen1.5-14B-ChatQwen1102
128Llama-3.2-3B-InstructMeta Llama1089
129granite-3.0-8b-instructIBM Granite1088
130Phi-3-small-8k-instructMicrosoft1086
131deepseek-llm-67b-chatDeepSeek1079
132Llama-2-70b-chat-hfMeta Llama1071
133Mistral-7B-Instruct-v0.2Mistral AI1060
134Qwen1.5-7B-ChatQwen1058
135Phi-3-mini-4k-instructMicrosoft1058
136gemma-1.1-7b-itGoogle1057
137granite-3.0-2b-instructIBM Granite1055
138SOLAR-10.7B-Instruct-v1.0Upstage1046
139Llama-2-13b-chat-hfMeta Llama1045
140Qwen-14B-ChatQwen1031
141Llama-3.2-1B-InstructMeta Llama1031
142Phi-3-mini-128k-instructMicrosoft1023
143gemma-7b-itGoogle1017
144Llama-2-7b-chat-hfMeta Llama1006
145gemma-1.1-2b-itGoogle992
146Qwen1.5-4B-ChatQwen978
147OLMo-7B-InstructAi2978
148gemma-2b-itGoogle970
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.