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)

147 models in creative writing ranked by LMArena score · updated Nov 11, 2026

#ModelScore
1GLM-5.2Z.ai1459
2GLM-5.3Z.ai1459
3GLM-5.1Z.ai1453
4Kimi-K3Moonshot AI1453
5DeepSeek-V4-ProDeepSeek1446
6GLM-5.3-FlashZ.ai1442
7Kimi-K3.1-ProMoonshot AI1439
8DeepSeek-V4.1-FlashDeepSeek1438
9MiMo-V2.5-ProXiaomi MiMo1437
10GLM-5Z.ai1437
11Kimi-K2.6Moonshot AI1433
12Kimi-K2.5Moonshot AI1422
13gemma-4-31B-itGoogle1415
14GLM-4.6Z.ai1412
15DeepSeek-R1-0528DeepSeek1405
16Hy3Tencent1405
17DeepSeek-V3.1-TerminusDeepSeek1403
18DeepSeek-V3.2-ExpDeepSeek1403
19DeepSeek-V4-FlashDeepSeek1403
20GLM-4.7Z.ai1402
21Qwen3.5-397B-A17BQwen1402
22gemma-4-26B-A4B-itGoogle1402
23DeepSeek-V3.1DeepSeek1401
24DeepSeek-V3.2DeepSeek1400
25NVIDIA-Nemotron-3-Ultra-550B-A55B-BF16NVIDIA1399
26InklingThinking Machines1395
27MiniMax-M3MiniMax1395
28GLM-4.5Z.ai1395
29MiMo-V2.5Xiaomi MiMo1391
30Beam-1.0-501B-A23Breflection1388
31Mistral-Large-3-675B-Instruct-2512Mistral AI1386
32Qwen3-235B-A22B-Thinking-2507Qwen1384
33NVIDIA-Nemotron-3.5-Super-VL-120B-A12B-BF16NVIDIA1377
34Qwen3.8-27BQwen1377
35MiMo-V2-FlashXiaomi MiMo1376
36Qwen3-235B-A22B-Instruct-2507Qwen1375
37Mistral-Medium-3.5-128BMistral AI1374
38Qwen3.5-122B-A10BQwen1368
39Qwen3-VL-235B-A22B-InstructQwen1365
40DeepSeek-V3-0324DeepSeek1364
41Qwen3.5-27BQwen1362
42Qwen3-235B-A22BQwen1357
43Step-3.5-FlashStepFun1357
44DeepSeek-R1DeepSeek1354
45MiniMax-M2.7MiniMax1354
46Inkling-2-SmallThinking Machines1348
47Qwen3.5-35B-A3BQwen1347
48GLM-4.6VZ.ai1346
49LongCat-Flash-ChatMeituan LongCat1345
50gemma-3-27b-itGoogle1345
51Qwen3-VL-235B-A22B-ThinkingQwen1344
52GLM-4.5-AirZ.ai1343
53Mistral-Large-4-1T-A52BMistral AI1343
54Muse-Glimmer-30Bmeta-models1341
55Hy3-previewTencent1340
56Inkling-SmallThinking Machines1337
57Qwen3-Next-80B-A3B-InstructQwen1333
58Qwen3-Coder-480B-A35B-InstructQwen1333
59gemma-3-12b-itGoogle1331
60MiniMax-M2.5MiniMax1330
61DeepSeek-V3DeepSeek1329
62Trinity-Large-PreviewArcee AI1319
63Qwen3-30B-A3B-Instruct-2507Qwen1317
64INTELLECT-3Prime Intellect1317
65NVIDIA-Nemotron-3-Super-120B-A12B-BF16NVIDIA1315
66Qwen3-Next-80B-A3B-ThinkingQwen1315
67Trinity-Large-ThinkingArcee AI1308
68Mistral-Small-3.2-24B-Instruct-2506Mistral AI1304
69Qwen3-32BQwen1297
70GLM-4.7-FlashZ.ai1296
71GLM-4.5VZ.ai1295
72Ling-flash-2.0inclusionAI1287
73QwQ-32BQwen1287
74MiniMax-M2MiniMax1287
75gemma-3n-E4B-itGoogle1287
76DeepSeek-V2.5-1210DeepSeek1285
77granite-4.2-30bIBM Granite1285
78gpt-oss-120bOpenAI1274
79Qwen3-30B-A3BQwen1271
80gemma-3-4b-itGoogle1270
81Ring-flash-2.0inclusionAI1270
82NVIDIA-Nemotron-3-Nano-30B-A3B-BF16NVIDIA1270
83Llama-3.1-Nemotron-70B-InstructNVIDIA1268
84Llama-4-Maverick-17B-128E-InstructMeta Llama1266
85Olmo-3.1-32B-InstructAi21266
86Llama-3.1-405B-InstructMeta Llama1262
87NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16NVIDIA1259
88Olmo-3-32B-ThinkAi21254
89Mistral-Small-3.1-24B-Instruct-2503Mistral AI1253
90granite-4.1-8bIBM Granite1251
91Llama-3.3-70B-InstructMeta Llama1250
92Llama-4-Scout-17B-16E-InstructMeta Llama1249
93gemma-2-27b-itGoogle1240
94DeepSeek-V2.5DeepSeek1237
95granite-4.2-8bIBM Granite1235
96Llama-3.1-70B-InstructMeta Llama1232
97Llama-3.1-Tulu-3-70BAi21231
98Olmo-3.1-32B-ThinkAi21227
99Qwen2.5-72B-InstructQwen1221
100granite-4.0-h-smallIBM Granite1210
101Meta-Llama-3-70B-InstructMeta Llama1210
102gemma-2-9b-itGoogle1205
103granite-4.2-3bIBM Granite1204
104Nemotron-4-340B-InstructNVIDIA1202
105gpt-oss-20bOpenAI1202
106aya-expanse-32bCohere Labs1200
107Mistral-Small-24B-Instruct-2501Mistral AI1195
108phi-4Microsoft1181
109Llama-3.1-Tulu-3-8BAi21181
110Qwen2-72B-InstructQwen1181
111Qwen2.5-Coder-32B-InstructQwen1174
112aya-expanse-8bCohere Labs1168
113Llama-3.1-8B-InstructMeta Llama1154
114Meta-Llama-3-8B-InstructMeta Llama1150
115Qwen1.5-110B-ChatQwen1148
116gemma-2-2b-itGoogle1147
117Mixtral-8x22B-Instruct-v0.1Mistral AI1141
118Qwen1.5-72B-ChatQwen1137
119QwQ-32B-PreviewQwen1134
120granite-3.1-8b-instructIBM Granite1129
121DeepSeek-Coder-V2-InstructDeepSeek1120
122granite-3.1-2b-instructIBM Granite1116
123Mixtral-8x7B-Instruct-v0.1Mistral AI1109
124Phi-3-medium-4k-instructMicrosoft1106
125Llama-3.2-3B-InstructMeta Llama1094
126Qwen1.5-14B-ChatQwen1090
127SOLAR-10.7B-Instruct-v1.0Upstage1089
128Qwen1.5-32B-ChatQwen1083
129Phi-3-small-8k-instructMicrosoft1083
130Llama-2-70b-chat-hfMeta Llama1075
131granite-3.0-8b-instructIBM Granite1071
132Mistral-7B-Instruct-v0.2Mistral AI1068
133deepseek-llm-67b-chatDeepSeek1067
134gemma-1.1-7b-itGoogle1059
135Llama-2-13b-chat-hfMeta Llama1047
136granite-3.0-2b-instructIBM Granite1046
137Phi-3-mini-4k-instructMicrosoft1044
138Qwen1.5-7B-ChatQwen1034
139Llama-2-7b-chat-hfMeta Llama1033
140Llama-3.2-1B-InstructMeta Llama1033
141Qwen-14B-ChatQwen1028
142Phi-3-mini-128k-instructMicrosoft1024
143gemma-7b-itGoogle1023
144gemma-1.1-2b-itGoogle998
145OLMo-7B-InstructAi2990
146gemma-2b-itGoogle986
147Qwen1.5-4B-ChatQwen969
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.