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 coding ranked by LMArena score · updated Nov 11, 2026

#ModelScore
1GLM-5.3-FlashZ.ai1509
2Kimi-K3Moonshot AI1506
3DeepSeek-V4.1-FlashDeepSeek1504
4MiMo-V2.5-ProXiaomi MiMo1501
5GLM-5.3Z.ai1499
6Kimi-K2.6Moonshot AI1489
7GLM-5.2Z.ai1488
8GLM-5.1Z.ai1485
9Kimi-K2.5Moonshot AI1484
10Kimi-K3.1-ProMoonshot AI1480
11Qwen3.8-27BQwen1479
12Beam-1.0-501B-A23Breflection1475
13MiMo-V2.5Xiaomi MiMo1469
14DeepSeek-V4-ProDeepSeek1469
15MiniMax-M3MiniMax1468
16NVIDIA-Nemotron-3-Ultra-550B-A55B-BF16NVIDIA1468
17Hy3Tencent1467
18InklingThinking Machines1467
19LongCat-Flash-ChatMeituan LongCat1467
20Qwen3.5-397B-A17BQwen1463
21GLM-5Z.ai1462
22Mistral-Medium-3.5-128BMistral AI1461
23gemma-4-31B-itGoogle1459
24Mistral-Large-4-1T-A52BMistral AI1457
25DeepSeek-V4-FlashDeepSeek1457
26Inkling-2-SmallThinking Machines1455
27GLM-4.7Z.ai1454
28MiniMax-M2.7MiniMax1453
29DeepSeek-V3.2DeepSeek1453
30Inkling-SmallThinking Machines1451
31Mistral-Large-3-675B-Instruct-2512Mistral AI1449
32GLM-4.6Z.ai1449
33gemma-4-26B-A4B-itGoogle1447
34Qwen3-235B-A22B-Instruct-2507Qwen1445
35MiMo-V2-FlashXiaomi MiMo1443
36Qwen3-Next-80B-A3B-InstructQwen1440
37DeepSeek-V3.2-ExpDeepSeek1439
38Qwen3-VL-235B-A22B-InstructQwen1438
39Qwen3.5-122B-A10BQwen1435
40Step-3.5-FlashStepFun1435
41GLM-4.5Z.ai1434
42NVIDIA-Nemotron-3.5-Super-VL-120B-A12B-BF16NVIDIA1433
43DeepSeek-R1-0528DeepSeek1427
44Qwen3.5-27BQwen1427
45DeepSeek-V3.1-TerminusDeepSeek1425
46Qwen3-VL-235B-A22B-ThinkingQwen1425
47Qwen3-235B-A22B-Thinking-2507Qwen1425
48Hy3-previewTencent1421
49Muse-Glimmer-30Bmeta-models1418
50DeepSeek-V3.1DeepSeek1417
51Qwen3-30B-A3B-Instruct-2507Qwen1416
52Qwen3-Coder-480B-A35B-InstructQwen1412
53Qwen3.5-35B-A3BQwen1410
54NVIDIA-Nemotron-3-Super-120B-A12B-BF16NVIDIA1404
55Qwen3-235B-A22BQwen1398
56GLM-4.5-AirZ.ai1397
57granite-4.2-30bIBM Granite1395
58Qwen3-Next-80B-A3B-ThinkingQwen1391
59GLM-4.6VZ.ai1391
60Ling-flash-2.0inclusionAI1389
61GLM-4.7-FlashZ.ai1382
62NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16NVIDIA1381
63Trinity-Large-PreviewArcee AI1381
64MiniMax-M2.5MiniMax1381
65gpt-oss-120bOpenAI1380
66NVIDIA-Nemotron-3-Nano-30B-A3B-BF16NVIDIA1378
67INTELLECT-3Prime Intellect1375
68granite-4.2-8bIBM Granite1374
69DeepSeek-R1DeepSeek1371
70MiniMax-M2MiniMax1371
71DeepSeek-V3-0324DeepSeek1368
72Ring-flash-2.0inclusionAI1366
73Mistral-Small-3.2-24B-Instruct-2506Mistral AI1363
74granite-4.2-3bIBM Granite1358
75Qwen3-32BQwen1358
76Trinity-Large-ThinkingArcee AI1357
77GLM-4.5VZ.ai1348
78Olmo-3.1-32B-InstructAi21347
79Qwen3-30B-A3BQwen1337
80QwQ-32BQwen1333
81DeepSeek-V3DeepSeek1325
82gemma-3-27b-itGoogle1322
83Olmo-3-32B-ThinkAi21318
84granite-4.1-8bIBM Granite1313
85DeepSeek-V2.5-1210DeepSeek1309
86Mistral-Small-3.1-24B-Instruct-2503Mistral AI1309
87gpt-oss-20bOpenAI1306
88Llama-4-Maverick-17B-128E-InstructMeta Llama1301
89DeepSeek-V2.5DeepSeek1301
90Qwen2.5-72B-InstructQwen1292
91Llama-3.1-405B-InstructMeta Llama1291
92Olmo-3.1-32B-ThinkAi21290
93Llama-4-Scout-17B-16E-InstructMeta Llama1286
94gemma-3-12b-itGoogle1281
95Qwen2.5-Coder-32B-InstructQwen1276
96Llama-3.1-Nemotron-70B-InstructNVIDIA1272
97gemma-3n-E4B-itGoogle1268
98Llama-3.3-70B-InstructMeta Llama1268
99Llama-3.1-70B-InstructMeta Llama1260
100DeepSeek-Coder-V2-InstructDeepSeek1251
101granite-4.0-h-smallIBM Granite1249
102Mistral-Small-24B-Instruct-2501Mistral AI1246
103Llama-3.1-Tulu-3-70BAi21235
104phi-4Microsoft1231
105gemma-3-4b-itGoogle1230
106gemma-2-27b-itGoogle1211
107Nemotron-4-340B-InstructNVIDIA1209
108Meta-Llama-3-70B-InstructMeta Llama1206
109aya-expanse-32bCohere Labs1196
110Qwen2-72B-InstructQwen1196
111Llama-3.1-8B-InstructMeta Llama1195
112granite-3.1-8b-instructIBM Granite1186
113Qwen1.5-110B-ChatQwen1184
114Llama-3.1-Tulu-3-8BAi21183
115gemma-2-9b-itGoogle1173
116Mixtral-8x22B-Instruct-v0.1Mistral AI1165
117Qwen1.5-72B-ChatQwen1165
118aya-expanse-8bCohere Labs1160
119QwQ-32B-PreviewQwen1155
120Qwen1.5-32B-ChatQwen1155
121Meta-Llama-3-8B-InstructMeta Llama1151
122granite-3.1-2b-instructIBM Granite1149
123Qwen1.5-14B-ChatQwen1138
124Phi-3-medium-4k-instructMicrosoft1130
125Mixtral-8x7B-Instruct-v0.1Mistral AI1126
126granite-3.0-8b-instructIBM Granite1112
127gemma-2-2b-itGoogle1112
128Qwen1.5-7B-ChatQwen1107
129Phi-3-small-8k-instructMicrosoft1101
130Llama-3.2-3B-InstructMeta Llama1098
131deepseek-llm-67b-chatDeepSeek1095
132Phi-3-mini-4k-instructMicrosoft1093
133granite-3.0-2b-instructIBM Granite1090
134gemma-1.1-7b-itGoogle1084
135Mistral-7B-Instruct-v0.2Mistral AI1082
136Llama-2-70b-chat-hfMeta Llama1079
137Qwen-14B-ChatQwen1071
138Llama-3.2-1B-InstructMeta Llama1070
139SOLAR-10.7B-Instruct-v1.0Upstage1067
140Llama-2-13b-chat-hfMeta Llama1062
141gemma-7b-itGoogle1048
142Phi-3-mini-128k-instructMicrosoft1039
143gemma-1.1-2b-itGoogle1034
144OLMo-7B-InstructAi21016
145gemma-2b-itGoogle1009
146Llama-2-7b-chat-hfMeta Llama1002
147Qwen1.5-4B-ChatQwen999
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.