Instruction Following
LMArena Text Leaderboard
Arena (arena.ai, formerly LMArena / Chatbot Arena) ranks models by millions of blind, head-to-head human votes on real prompts. The Arena Score shown here is a Bradley-Terry rating computed with Arena's default style control, so higher means people preferred that model's answers more often — it's a relative rating, not a percentage, so compare it only to other models on this board.
Source: lmarena182 open models ranked+209 proprietaryData through Sep 2026
Open models ranked on LMArena Text
# shows rank among open models / rank overall (including proprietary).
| # | Model | Score |
|---|---|---|
| 1 / 17 | Kimi K3 · 2779.9B | 1484.8 |
| 2 / 19 | GLM 5.3 · 753.3B | 1483.0 |
| 3 / 29 | GLM 5.3 Flash · 321.3B | 1475.4 |
| 4 / 38 | GLM 5.2 · 753.3B | 1472.1 |
| 5 / 44 | MiMo V2.5 Pro · 1023.2B | 1467.4 |
| 6 / 47 | GLM 5.1 · 753.9B | 1465.5 |
| 7 / 50 | DeepSeek V4 Pro 0813 · 1650.5B | 1463.4 |
| 8 / 52 | Kimi K2.6 · 1026.9B | 1460.4 |
| 9 / 56 | GLM 5 · 753.9B | 1457.6 |
| 10 / 57 | DeepSeek V4 Pro · 1598.8B | 1457.3 |
| 11 / 60 | Hy3 · 298.8B | 1455.8 |
| 12 / 67 | Gemma 4 31B IT · 31.3B | 1451.1 |
| 13 / 68 | Kimi K2.5 · 1026.9B | 1450.4 |
| 14 / 81 | Qwen3.5 397B A17B · 403.4B | 1441.8 |
| 15 / 82 | GLM 4.7 · 358.3B | 1441.7 |
| 16 / 83 | MiniMax M3 · 427.0B | 1441.3 |
| 17 / 84 | Inkling · 952.4B | 1440.1 |
| 18 / 86 | DeepSeek V4 Flash · 290.9B | 1437.9 |
| 19 / 87 | Gemma 4 26B A4B IT · 25.8B | 1437.9 |
| 20 / 89 | Qwen3.8 27B · 27.8B | 1437.1 |
| 21 / 94 | MiMo V2.5 · 310.8B | 1433.6 |
| 22 / 99 | Kimi K2 Thinking · 1026.4B | 1430.1 |
| 23 / 101 | Muse Glimmer 30B · 29.8B | 1427.2 |
| 24 / 103 | Mistral Medium 3.5 128B · 127.7B | 1426.3 |
| 25 / 106 | NVIDIA Nemotron 3 Ultra 550B A55B BF16 · 560.5B | 1425.6 |
| 26 / 107 | DeepSeek V3.2 Exp · 685.4B | 1425.3 |
| 27 / 108 | DeepSeek V3.2 · 685.4B | 1425.2 |
| 28 / 109 | GLM 4.6 · 356.8B | 1424.7 |
| 29 / 111 | Qwen3 235B A22B Instruct 2507 · 235.1B | 1423.0 |
| 30 / 112 | DeepSeek R1 0528 · 684.5B | 1421.4 |
| 31 / 115 | Kimi K2 Instruct 0905 · 1026.5B | 1418.1 |
| 32 / 116 | Kimi K2 Instruct · 1026.4B | 1417.8 |
| 33 / 117 | DeepSeek V3.1 Terminus · 684.5B | 1417.4 |
| 34 / 118 | DeepSeek V3.1 · 684.5B | 1417.3 |
| 35 / 119 | Qwen3.5 122B A10B · 125.1B | 1416.8 |
| 36 / 120 | MiniMax M2.7 · 228.7B | 1415.2 |
| 37 / 124 | Qwen3 VL 235B A22B Instruct · 235.7B | 1414.1 |
| 38 / 126 | Mistral Large 3 675B Instruct 2512 · 675B | 1413.0 |
| 39 / 127 | Hy3 Preview · 298.8B | 1412.7 |
| 40 / 129 | GLM 4.5 · 358.3B | 1411.3 |
| 41 / 133 | Qwen3.5 27B · 27.8B | 1407.9 |
| 42 / 134 | Inkling Small · 266.0B | 1404.6 |
| 43 / 137 | Qwen3 235B A22B · 235.1B | 1402.5 |
| 44 / 140 | LongCat Flash Chat · 561.9B | 1401.3 |
| 45 / 142 | Qwen3 235B A22B Thinking 2507 · 235.1B | 1400.1 |
| 46 / 143 | Qwen3 Next 80B A3B Instruct · 81.3B | 1399.3 |
| 47 / 144 | DeepSeek R1 · 684.5B | 1398.0 |
| 48 / 146 | DeepSeek v3 0324 · 684.5B | 1395.7 |
| 49 / 147 | Qwen3 VL 235B A22B Thinking · 235.7B | 1395.2 |
| 50 / 149 | Qwen3.5 35B A3B · 36.0B | 1395.0 |
| 51 / 151 | Step 3.5 Flash · 199.4B | 1394.1 |
| 52 / 152 | MiMo v2 Flash · 309.8B | 1391.9 |
| 53 / 155 | MiniMax M2.5 · 228.7B | 1390.3 |
| 54 / 159 | Qwen3 Coder 480B A35B Instruct · 480.2B | 1387.5 |
| 55 / 162 | MiniMax M2.1 · 228.7B | 1383.8 |
| 56 / 165 | Qwen3 30B A3B Instruct 2507 · 30.5B | 1382.3 |
| 57 / 167 | GLM 4.6V · 107.7B | 1378.5 |
| 58 / 168 | Trinity Large Preview · 398.6B | 1378.4 |
| 59 / 173 | GLM 4.5 Air · 110.5B | 1373.2 |
| 60 / 175 | Qwen3 Next 80B A3B Thinking · 81.3B | 1369.0 |
| 61 / 176 | Trinity Large Thinking · 398.6B | 1369.0 |
| 62 / 177 | GLM 4.7 Flash · 31.2B | 1366.3 |
| 63 / 178 | Gemma 3 27B IT · 27.4B | 1365.2 |
| 64 / 183 | NVIDIA Nemotron 3 Super 120B A12B BF16 · 123.6B | 1360.5 |
| 65 / 185 | DeepSeek v3 · 684.5B | 1358.4 |
| 66 / 187 | Mistral Small 3.2 24B Instruct 2506 · 24.0B | 1356.7 |
| 67 / 188 | INTELLECT 3 · 106.9B | 1356.3 |
| 68 / 189 | C4ai Command A 03 2025 · 111.1B | 1353.6 |
| 69 / 191 | GLM 4.5V · 107.7B | 1352.6 |
| 70 / 192 | GPT OSS 120B · 116.8B | 1352.3 |
| 71 / 193 | NVIDIA Nemotron 3.5 Lightning 30B A3B BF16 · 31.6B | 1351.5 |
| 72 / 196 | Step3 · 321.0B | 1349.4 |
| 73 / 199 | Llama 3 1 Nemotron Ultra 253B V1 · 253.4B | 1347.5 |
| 74 / 200 | Qwen3 32B · 32.8B | 1346.9 |
| 75 / 205 | MiniMax M2 · 228.7B | 1345.9 |
| 76 / 206 | Ling Flash 2.0 · 102.9B | 1344.0 |
| 77 / 207 | Llama 3 3 Nemotron Super 49B V1 5 · 49.9B | 1343.1 |
| 78 / 210 | Granite 4.2 30B · 29.3B | 1341.9 |
| 79 / 211 | Gemma 3 12B IT · 12.2B | 1341.7 |
| 80 / 216 | QwQ 32B · 32.8B | 1335.8 |
| 81 / 220 | Llama 3.1 405B Instruct · 405.9B | 1335.2 |
| 82 / 222 | Olmo 3.1 32B Instruct · 32.2B | 1330.0 |
| 83 / 224 | Molmo2 8B · 8.7B | 1328.0 |
| 84 / 225 | Llama 3 3 Nemotron Super 49B V1 · 49.9B | 1327.7 |
| 85 / 226 | Llama 4 Maverick 17B 128E Instruct · 401.6B | 1326.9 |
| 86 / 227 | Qwen3 30B A3B · 30.5B | 1326.9 |
| 87 / 232 | DeepSeek V2.5 1210 · 235.7B | 1323.3 |
| 88 / 235 | Llama 4 Scout 17B 16E Instruct · 108.6B | 1321.5 |
| 89 / 236 | Ring Flash 2.0 · 102.9B | 1320.9 |
| 90 / 240 | Llama 3.3 70B Instruct · 70.6B | 1317.6 |
| 91 / 242 | GPT OSS 20B · 20.9B | 1317.2 |
| 92 / 243 | Gemma 3n E4B IT · 7.8B | 1317.2 |
| 93 / 245 | NVIDIA Nemotron 3 Nano 30B A3B BF16 · 31.6B | 1314.2 |
| 94 / 246 | Mistral Large Instruct 2407 · 122.6B | 1314.2 |
| 95 / 247 | Athene v2 Chat · 72.7B | 1314.1 |
| 96 / 254 | DeepSeek V2.5 · 235.7B | 1307.1 |
| 97 / 255 | Olmo 3 32B Think · 32.2B | 1306.7 |
| 98 / 257 | Granite 4.1 8B · 8.8B | 1306.0 |
| 99 / 258 | Mistral Large Instruct 2411 · 122.6B | 1305.4 |
| 100 / 260 | Gemma 3 4B IT · 4.3B | 1303.1 |
| 101 / 261 | Mistral Small 3.1 24B Instruct 2503 · 24.0B | 1303.1 |
| 102 / 262 | Qwen2.5 72B Instruct · 72.7B | 1302.8 |
| 103 / 263 | Llama 3.1 Nemotron 70B Instruct HF · 70.6B | 1298.7 |
| 104 / 265 | Granite 4.2 3B · 3.7B | 1293.7 |
| 105 / 266 | Llama 3.1 70B Instruct · 70.6B | 1293.1 |
| 106 / 268 | AI21 Jamba Large 1.5 · 398.6B | 1289.3 |
| 107 / 269 | Gemma 2 27B IT · 27.2B | 1289.2 |
| 108 / 272 | Granite 4.2 8B · 8.8B | 1286.6 |
| 109 / 273 | Llama 3 1 Nemotron 51B Instruct · 51B | 1286.3 |
| 110 / 275 | Llama 3.1 Tulu 3 70B · 70.6B | 1285.9 |
| 111 / 276 | Granite 4.0 H Small · 32.2B | 1285.5 |
| 112 / 277 | Olmo 3.1 32B Think · 32.2B | 1285.5 |
| 113 / 279 | Gemma 2 9B IT SimPO · 9.2B | 1280.1 |
| 114 / 281 | Meta Llama 3 70B Instruct · 70.6B | 1276.3 |
| 115 / 282 | C4ai Command R Plus 08 2024 · 103.8B | 1276.1 |
| 116 / 284 | Mistral Small 24B Instruct 2501 · 23.6B | 1274.3 |
| 117 / 287 | Qwen2.5 Coder 32B Instruct · 32.8B | 1270.5 |
| 118 / 288 | Aya Expanse 32B · 32.3B | 1267.0 |
| 119 / 289 | Gemma 2 9B IT · 9.2B | 1266.7 |
| 120 / 290 | DeepSeek Coder v2 Instruct · 235.7B | 1265.1 |
| 121 / 291 | Qwen2 72B Instruct · 72.7B | 1261.5 |
| 122 / 292 | C4ai Command R Plus · 103.8B | 1261.5 |
| 123 / 296 | Phi 4 · 14.7B | 1256.1 |
| 124 / 297 | OLMo 2 0325 32B Instruct · 32.2B | 1251.5 |
| 125 / 298 | C4ai Command R 08 2024 · 32.3B | 1250.1 |
| 126 / 301 | AI21 Jamba Mini 1.5 · 51.6B | 1239.6 |
| 127 / 302 | Ministral 8B Instruct 2410 · 8.0B | 1237.5 |
| 128 / 304 | Qwen1.5 110B Chat · 111.2B | 1233.9 |
| 129 / 306 | Qwen1.5 72B Chat · 72.3B | 1233.2 |
| 130 / 308 | Mixtral 8x22B Instruct v0.1 · 140.6B | 1229.3 |
| 131 / 309 | C4ai Command R V01 · 35.0B | 1226.6 |
| 132 / 312 | Meta Llama 3 8B Instruct · 8.0B | 1223.4 |
| 133 / 314 | Aya Expanse 8B · 8.0B | 1222.9 |
| 134 / 316 | Llama 3.1 Tulu 3 8B · 8.0B | 1220.3 |
| 135 / 317 | Zephyr Orpo 141B A35b v0.1 · 140.6B | 1212.7 |
| 136 / 318 | Yi 1.5 34B Chat · 34.4B | 1212.6 |
| 137 / 319 | Llama 3.1 8B Instruct · 8.0B | 1211.1 |
| 138 / 320 | Granite 3.1 8B Instruct · 8.2B | 1208.2 |
| 139 / 322 | Qwen1.5 32B Chat · 32.5B | 1203.7 |
| 140 / 323 | Gemma 2 2B IT · 2.6B | 1199.9 |
| 141 / 324 | Phi 3 Medium 4k Instruct · 14.0B | 1197.6 |
| 142 / 325 | Mixtral 8x7B Instruct v0.1 · 46.7B | 1196.9 |
| 143 / 327 | Qwen1.5 14B Chat · 14.2B | 1190.8 |
| 144 / 328 | Internlm2 5 20B Chat · 19.9B | 1190.6 |
| 145 / 329 | Deepseek Llm 67B Chat · 67B | 1184.6 |
| 146 / 331 | Yi 34B Chat · 34.4B | 1183.5 |
| 147 / 332 | Granite 3.0 8B Instruct · 8.2B | 1182.8 |
| 148 / 334 | Openchat 3.5 0106 · 7.2B | 1182.4 |
| 149 / 335 | Gemma 1.1 7B IT · 8.5B | 1182.2 |
| 150 / 336 | Snowflake Arctic Instruct · 478.6B | 1179.7 |
| 151 / 337 | Granite 3.1 2B Instruct · 2.5B | 1178.6 |
| 152 / 339 | OpenHermes 2.5 Mistral 7B · 7B | 1175.6 |
| 153 / 340 | Vicuna 33B V1.3 · 33B | 1172.6 |
| 154 / 341 | Starling LM 7B Beta · 7.2B | 1170.8 |
| 155 / 342 | Phi 3 Small 8k Instruct · 7.4B | 1170.8 |
| 156 / 343 | Llama 2 70B Chat HF · 69.0B | 1170.4 |
| 157 / 344 | Starling LM 7B Alpha · 7.2B | 1167.1 |
| 158 / 345 | Llama 3.2 3B Instruct · 3.2B | 1166.5 |
| 159 / 346 | Nous Hermes 2 Mixtral 8x7B DPO · 46.7B | 1164.2 |
| 160 / 347 | Granite 3.0 2B Instruct · 2.6B | 1156.3 |
| 161 / 349 | QwQ 32B Preview · 32.8B | 1154.2 |
| 162 / 350 | SOLAR 10.7B Instruct v1.0 · 10.7B | 1152.1 |
| 163 / 354 | Mistral 7B Instruct v0.2 · 7.2B | 1149.1 |
| 164 / 356 | Qwen1.5 7B Chat · 7.7B | 1143.6 |
| 165 / 357 | Phi 3 Mini 4k Instruct · 3.8B | 1142.8 |
| 166 / 359 | Llama 2 13B Chat HF · 13.0B | 1141.2 |
| 167 / 360 | Qwen 14B Chat · 14.2B | 1139.0 |
| 168 / 362 | Gemma 7B IT · 8.5B | 1137.4 |
| 169 / 363 | CodeLlama 34B Instruct HF · 33.7B | 1136.6 |
| 170 / 364 | Zephyr 7B Beta · 7.2B | 1130.5 |
| 171 / 365 | Phi 3 Mini 128k Instruct · 3.8B | 1129.5 |
| 172 / 368 | StripedHyena Nous 7B · 7.6B | 1121.3 |
| 173 / 369 | CodeLlama 70B Instruct HF · 69.0B | 1118.9 |
| 174 / 370 | Gemma 1.1 2B IT · 2.5B | 1116.3 |
| 175 / 372 | SmolLM2 1.7B Instruct · 1.7B | 1114.4 |
| 176 / 373 | Llama 3.2 1B Instruct · 1.2B | 1110.8 |
| 177 / 374 | Mistral 7B Instruct v0.1 · 7.2B | 1110.0 |
| 178 / 375 | Llama 2 7B Chat HF · 6.7B | 1107.7 |
| 179 / 376 | Gemma 2B IT · 2.5B | 1093.3 |
| 180 / 383 | Chatglm3 6B · 6.2B | 1056.2 |
| 181 / 385 | Chatglm2 6B · 6B | 1024.4 |
| 182 / 391 | Stablelm Tuned Alpha 7B · 7B | 952.9 |
Score vs model size
Which models give the most quality for their size — the ones worth running locally.
- Llama 3.2 1B Instruct, 1B, score 1110.8 — on the efficiency frontier (best score at its size or smaller).
- SmolLM2 1.7B Instruct, 2B, score 1114.4 — on the efficiency frontier (best score at its size or smaller).
- Gemma 1.1 2B IT, 3B, score 1116.3 — on the efficiency frontier (best score at its size or smaller).
- Granite 3.1 2B Instruct, 3B, score 1178.6 — on the efficiency frontier (best score at its size or smaller).
- Gemma 2 2B IT, 3B, score 1199.9 — on the efficiency frontier (best score at its size or smaller).
- Granite 4.2 3B, 4B, score 1293.7 — on the efficiency frontier (best score at its size or smaller).
- Gemma 3 4B IT, 4B, score 1303.1 — on the efficiency frontier (best score at its size or smaller).
- Gemma 3n E4B IT, 8B, score 1317.2 — on the efficiency frontier (best score at its size or smaller).
- Molmo2 8B, 9B, score 1328.0 — on the efficiency frontier (best score at its size or smaller).
- Gemma 3 12B IT, 12B, score 1341.7 — on the efficiency frontier (best score at its size or smaller).
- Mistral Small 3.2 24B Instruct 2506, 24B, score 1356.7 — on the efficiency frontier (best score at its size or smaller).
- Gemma 4 26B A4B IT, 26B, score 1437.9 — on the efficiency frontier (best score at its size or smaller).
- Gemma 4 31B IT, 31B, score 1451.1 — on the efficiency frontier (best score at its size or smaller).
- Hy3, 299B, score 1455.8 — on the efficiency frontier (best score at its size or smaller).
- GLM 5.3 Flash, 321B, score 1475.4 — on the efficiency frontier (best score at its size or smaller).
- GLM 5.3, 753B, score 1483.0 — on the efficiency frontier (best score at its size or smaller).
- Kimi K3, 2.8T, score 1484.8 — on the efficiency frontier (best score at its size or smaller).
LMArena Text: frequently asked questions
- What is the best open LLM on LMArena Text?
- Kimi K3 is the top open model on LMArena Text, scoring 1484.8. Among all models tested — including proprietary ones — it ranks #17. The top model overall is Claude Fable 5 (Anthropic) at 1505.7.
- What's the best LMArena Text model you can run on a 24 GB GPU?
- Gemma 4 31B IT is the highest-scoring open model that fits in 24 GB at 4-bit quantization (about 17 GB), scoring 1451.1 on LMArena Text.
- What's the best LMArena Text model you can run on a 12 GB GPU?
- Gemma 3 12B IT is the highest-scoring open model that fits in 12 GB at 4-bit quantization (about 7 GB), scoring 1341.7 on LMArena Text.
- Can open models match proprietary models on LMArena Text?
- Not quite on LMArena Text: the strongest proprietary model (Claude Fable 5) scores 1505.7, ahead of the best open model (Kimi K3) at 1484.8 — but you can run the open one yourself.
Scores aggregated from lmarena. llmrun does not run this benchmark — see the source for methodology, or the about benchmarks for what it measures.