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

147 models in multi-turn ranked by LMArena score · updated Nov 11, 2026

#ModelScore
1Kimi-K3Moonshot AI1489
2MiMo-V2.5-ProXiaomi MiMo1476
3GLM-5.3Z.ai1474
4GLM-5.1Z.ai1472
5GLM-5.3-FlashZ.ai1471
6GLM-5.2Z.ai1469
7DeepSeek-V4-ProDeepSeek1467
8DeepSeek-V4.1-FlashDeepSeek1462
9GLM-5Z.ai1456
10Kimi-K2.6Moonshot AI1454
11gemma-4-31B-itGoogle1452
12DeepSeek-V4-FlashDeepSeek1449
13Kimi-K3.1-ProMoonshot AI1449
14GLM-4.7Z.ai1446
15MiMo-V2.5Xiaomi MiMo1445
16Qwen3.5-397B-A17BQwen1445
17Kimi-K2.5Moonshot AI1444
18InklingThinking Machines1442
19gemma-4-26B-A4B-itGoogle1442
20MiniMax-M3MiniMax1439
21Hy3Tencent1438
22Qwen3.8-27BQwen1436
23Qwen3-235B-A22B-Instruct-2507Qwen1432
24Mistral-Large-3-675B-Instruct-2512Mistral AI1429
25Qwen3-VL-235B-A22B-InstructQwen1428
26GLM-4.6Z.ai1427
27DeepSeek-V3.2DeepSeek1427
28Mistral-Medium-3.5-128BMistral AI1424
29DeepSeek-V3.2-ExpDeepSeek1421
30Beam-1.0-501B-A23Breflection1421
31Mistral-Large-4-1T-A52BMistral AI1421
32Inkling-SmallThinking Machines1417
33Qwen3.5-122B-A10BQwen1416
34NVIDIA-Nemotron-3.5-Super-VL-120B-A12B-BF16NVIDIA1415
35Qwen3-Next-80B-A3B-InstructQwen1415
36GLM-4.5Z.ai1415
37MiniMax-M2.7MiniMax1412
38DeepSeek-V3.1-TerminusDeepSeek1410
39Hy3-previewTencent1410
40Qwen3.5-27BQwen1410
41NVIDIA-Nemotron-3-Ultra-550B-A55B-BF16NVIDIA1410
42Inkling-2-SmallThinking Machines1408
43DeepSeek-V3.1DeepSeek1407
44Qwen3-235B-A22B-Thinking-2507Qwen1407
45DeepSeek-R1-0528DeepSeek1405
46Step-3.5-FlashStepFun1404
47MiMo-V2-FlashXiaomi MiMo1404
48LongCat-Flash-ChatMeituan LongCat1403
49Qwen3-235B-A22BQwen1398
50Muse-Glimmer-30Bmeta-models1397
51DeepSeek-R1DeepSeek1390
52Qwen3.5-35B-A3BQwen1390
53DeepSeek-V3-0324DeepSeek1389
54Qwen3-VL-235B-A22B-ThinkingQwen1388
55Qwen3-30B-A3B-Instruct-2507Qwen1378
56GLM-4.5-AirZ.ai1371
57NVIDIA-Nemotron-3-Super-120B-A12B-BF16NVIDIA1368
58Qwen3-Coder-480B-A35B-InstructQwen1365
59MiniMax-M2.5MiniMax1364
60MiniMax-M2MiniMax1361
61GLM-4.6VZ.ai1358
62INTELLECT-3Prime Intellect1350
63DeepSeek-V3DeepSeek1349
64Qwen3-Next-80B-A3B-ThinkingQwen1346
65gemma-3-27b-itGoogle1345
66Mistral-Small-3.2-24B-Instruct-2506Mistral AI1343
67GLM-4.7-FlashZ.ai1341
68Trinity-Large-PreviewArcee AI1341
69granite-4.2-30bIBM Granite1340
70gpt-oss-120bOpenAI1340
71gemma-3-12b-itGoogle1334
72GLM-4.5VZ.ai1333
73Qwen3-32BQwen1331
74NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16NVIDIA1324
75Ling-flash-2.0inclusionAI1323
76Trinity-Large-ThinkingArcee AI1321
77NVIDIA-Nemotron-3-Nano-30B-A3B-BF16NVIDIA1319
78QwQ-32BQwen1313
79Olmo-3.1-32B-InstructAi21310
80Qwen3-30B-A3BQwen1307
81granite-4.2-8bIBM Granite1298
82Llama-3.1-405B-InstructMeta Llama1297
83DeepSeek-V2.5-1210DeepSeek1297
84Olmo-3-32B-ThinkAi21290
85Llama-4-Maverick-17B-128E-InstructMeta Llama1289
86granite-4.2-3bIBM Granite1288
87Ring-flash-2.0inclusionAI1287
88Llama-4-Scout-17B-16E-InstructMeta Llama1280
89Llama-3.3-70B-InstructMeta Llama1279
90gemma-3n-E4B-itGoogle1276
91Llama-3.1-Nemotron-70B-InstructNVIDIA1275
92Qwen2.5-72B-InstructQwen1272
93Mistral-Small-3.1-24B-Instruct-2503Mistral AI1271
94granite-4.1-8bIBM Granite1268
95gpt-oss-20bOpenAI1268
96DeepSeek-V2.5DeepSeek1261
97Llama-3.1-70B-InstructMeta Llama1256
98gemma-3-4b-itGoogle1255
99Olmo-3.1-32B-ThinkAi21252
100Llama-3.1-Tulu-3-70BAi21251
101granite-4.0-h-smallIBM Granite1243
102gemma-2-27b-itGoogle1224
103Meta-Llama-3-70B-InstructMeta Llama1223
104Qwen2.5-Coder-32B-InstructQwen1222
105Mistral-Small-24B-Instruct-2501Mistral AI1217
106Nemotron-4-340B-InstructNVIDIA1214
107phi-4Microsoft1206
108Qwen2-72B-InstructQwen1195
109gemma-2-9b-itGoogle1193
110aya-expanse-32bCohere Labs1190
111DeepSeek-Coder-V2-InstructDeepSeek1176
112Llama-3.1-8B-InstructMeta Llama1172
113Qwen1.5-72B-ChatQwen1160
114Qwen1.5-110B-ChatQwen1160
115aya-expanse-8bCohere Labs1160
116Llama-3.1-Tulu-3-8BAi21154
117Meta-Llama-3-8B-InstructMeta Llama1152
118QwQ-32B-PreviewQwen1141
119Qwen1.5-32B-ChatQwen1140
120Mixtral-8x22B-Instruct-v0.1Mistral AI1130
121gemma-2-2b-itGoogle1118
122Mixtral-8x7B-Instruct-v0.1Mistral AI1115
123Qwen1.5-14B-ChatQwen1110
124granite-3.1-8b-instructIBM Granite1108
125Llama-3.2-3B-InstructMeta Llama1105
126granite-3.1-2b-instructIBM Granite1099
127Llama-2-70b-chat-hfMeta Llama1088
128Phi-3-medium-4k-instructMicrosoft1088
129deepseek-llm-67b-chatDeepSeek1082
130Phi-3-small-8k-instructMicrosoft1068
131granite-3.0-8b-instructIBM Granite1063
132Mistral-7B-Instruct-v0.2Mistral AI1062
133Qwen1.5-7B-ChatQwen1061
134granite-3.0-2b-instructIBM Granite1053
135Llama-2-13b-chat-hfMeta Llama1050
136Phi-3-mini-4k-instructMicrosoft1049
137SOLAR-10.7B-Instruct-v1.0Upstage1045
138gemma-1.1-7b-itGoogle1040
139Llama-3.2-1B-InstructMeta Llama1030
140Llama-2-7b-chat-hfMeta Llama1029
141Qwen-14B-ChatQwen1022
142OLMo-7B-InstructAi21007
143Phi-3-mini-128k-instructMicrosoft989
144Qwen1.5-4B-ChatQwen977
145gemma-7b-itGoogle963
146gemma-1.1-2b-itGoogle958
147gemma-2b-itGoogle945
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.