The best free AI models, ranked by real votes
Only models you can use without paying: free online, or downloaded to your own GPU. Scores come from millions of blind human votes on Arena.
226 of 416 models in Arena's Chat ranking are free to use. Votes as of 2026-10-15. Best free: mimo-v2.6-pro, #11 overall.
| Free rank | Model | Arena score | Use free online | Run at home |
|---|---|---|---|---|
| 1 | mimo-v2.6-pro xiaomi · MIT |
1491 #11 of 416 |
— | 1024.2 B~678 GB at 4-bit · server: 9 GPUs of 80 GB |
| 2 | Kimi K3 (Max) moonshot · Kimi K3 license |
1475 #25 of 416 |
Lab chat | 2779.9 B~1837 GB at 4-bit · server: 23 GPUs of 80 GB |
| 3 | glm-5.3-max zai · MIT |
1471 #27 of 416 |
Lab chat | 753.3 B~499 GB at 4-bit · server: 7 GPUs of 80 GB |
| 4 | glm-5.3-flash zai · MIT |
1471 #29 of 416 |
Lab chat | 321.3 B~214 GB at 4-bit · 9 GPUs of 24 GB |
| 5 | glm-5.2-max zai · MIT |
1470 #31 of 416 |
Lab chat | 753.3 B~499 GB at 4-bit · server: 7 GPUs of 80 GB |
| 6 | mimo-v2.5-pro xiaomi · MIT |
1465 #37 of 416 |
— | 1023.2 B~677 GB at 4-bit · server: 9 GPUs of 80 GB |
| 7 | deepseek-v4.1-flash-max deepseek · MIT |
1463 #38 of 416 |
Lab chat | 763.2 B~506 GB at 4-bit · server: 7 GPUs of 80 GB |
| 8 | glm-5.1 zai · MIT |
1462 #39 of 416 |
Lab chat | 753.9 B~500 GB at 4-bit · server: 7 GPUs of 80 GB |
| 9 | kimi-k2.6 moonshot · Modified MIT |
1455 #44 of 416 |
Lab chat | 1026.9 B~680 GB at 4-bit · server: 9 GPUs of 80 GB |
| 10 | mimo-v2.6-flash xiaomi · MIT |
1455 #45 of 416 |
— | 310.8 B~207 GB at 4-bit · 9 GPUs of 24 GB |
| 11 | deepseek-v4-pro deepseek · MIT |
1451 #49 of 416 |
Lab chat | 1598.8 B~1057 GB at 4-bit · server: 14 GPUs of 80 GB |
| 12 | glm-5 zai · MIT |
1446 #60 of 416 |
Lab chat | 753.9 B~500 GB at 4-bit · server: 7 GPUs of 80 GB |
| 13 | deepseek-v4-pro-high-20260813 deepseek · MIT |
1446 #61 of 416 |
Lab chat | Size unknown |
| 14 | kimi-k2.5-thinking moonshot · Modified MIT |
1445 #62 of 416 |
Lab chat | Size unknown |
| 15 | deepseek-v4-pro-high-preview deepseek · MIT |
1445 #63 of 416 |
Lab chat | Size unknown |
| 16 | nvidia-nemotron-3-ultra-550b-a55b-nvfp4 nvidia · OpenMDW-1.1 |
1445 #64 of 416 |
Lab chat | 560.5 B~372 GB at 4-bit · server: 5 GPUs of 80 GB |
| 17 | gemma-4-31b google · Apache 2.0 |
1443 #67 of 416 |
Free API · try itLab chat | 31.3 B~23 GB at 4-bit · 1 GPU of 24 GB |
| 18 | inkling thinky · Apache 2.0 |
1441 #74 of 416 |
Free API · try it | Size unknown |
| 19 | qwen3.8-27b alibaba · Apache 2.0 |
1441 #75 of 416 |
Lab chat | 27.8 B~20 GB at 4-bit · 1 GPU of 24 GB |
| 20 | glm-4.6 zai · MIT |
1440 #76 of 416 |
Lab chat | 356.8 B~237 GB at 4-bit · 10 GPUs of 24 GB |
| 21 | hy3 tencent · Apache 2.0 |
1438 #80 of 416 |
Lab chat | 298.8 B~199 GB at 4-bit · 9 GPUs of 24 GB |
| 22 | qwen3.5-397b-a17b alibaba · Apache 2.0 |
1438 #81 of 416 |
Lab chat | 403.4 B~268 GB at 4-bit · server: 4 GPUs of 80 GB |
| 23 | glm-4.7 zai · MIT |
1435 #87 of 416 |
Lab chat | 358.3 B~238 GB at 4-bit · 10 GPUs of 24 GB |
| 24 | gemma-4-26b-a4b google · Apache 2.0 |
1434 #90 of 416 |
Free API · try itLab chat | 25.8 B~19 GB at 4-bit · 1 GPU of 24 GB |
| 25 | minimax-m3 minimax · MiniMax Community License |
1433 #91 of 416 |
Lab chat | 427 B~284 GB at 4-bit · server: 4 GPUs of 80 GB |
| 26 | deepseek-v4-flash deepseek · MIT |
1432 #92 of 416 |
Lab chat | 290.9 B~194 GB at 4-bit · 9 GPUs of 24 GB |
| 27 | glm-4.5 zai · MIT |
1430 #95 of 416 |
Lab chat | 358.3 B~238 GB at 4-bit · 10 GPUs of 24 GB |
| 28 | mistral-large-3 mistral · Apache 2.0 |
1428 #98 of 416 |
— | Size unknown |
| 29 | mimo-v2.5 xiaomi · MIT |
1428 #100 of 416 |
— | 310.8 B~207 GB at 4-bit · 9 GPUs of 24 GB |
| 30 | deepseek-r1-0528 deepseek · MIT |
1428 #101 of 416 |
Lab chat | 684.5 B~454 GB at 4-bit · server: 6 GPUs of 80 GB |
| 31 | deepseek-v3.2-exp-thinking deepseek · MIT |
1425 #104 of 416 |
Lab chat | 685.4 B~454 GB at 4-bit · server: 6 GPUs of 80 GB |
| 32 | deepseek-v3.2 deepseek · MIT |
1424 #105 of 416 |
Lab chat | 685.4 B~454 GB at 4-bit · server: 6 GPUs of 80 GB |
| 33 | deepseek-v4-flash-high-preview deepseek · MIT |
1424 #106 of 416 |
Lab chat | Size unknown |
| 34 | longcat-flash-chat meituan · MIT |
1423 #109 of 416 |
— | 561.9 B~373 GB at 4-bit · server: 5 GPUs of 80 GB |
| 35 | deepseek-v3.2-exp deepseek · MIT |
1422 #111 of 416 |
Lab chat | 685.4 B~454 GB at 4-bit · server: 6 GPUs of 80 GB |
| 36 | mistral-medium-3.5 mistral · Modified MIT |
1421 #112 of 416 |
— | Size unknown |
| 37 | kimi-k2.5-instant moonshot · Modified MIT |
1420 #113 of 416 |
Lab chat | Size unknown |
| 38 | qwen3-vl-235b-a22b-instruct alibaba · Apache 2.0 |
1420 #114 of 416 |
Lab chat | 235.7 B~158 GB at 4-bit · 7 GPUs of 24 GB |
| 39 | deepseek-v3.2-thinking deepseek · MIT |
1420 #115 of 416 |
Lab chat | 685.4 B~454 GB at 4-bit · server: 6 GPUs of 80 GB |
| 40 | deepseek-v3.1 deepseek · MIT |
1420 #116 of 416 |
Lab chat | 684.5 B~454 GB at 4-bit · server: 6 GPUs of 80 GB |
| 41 | qwen3-235b-a22b-instruct-2507 alibaba · Apache 2.0 |
1419 #119 of 416 |
Lab chat | 235.1 B~157 GB at 4-bit · 7 GPUs of 24 GB |
| 42 | deepseek-v3.1-terminus-thinking deepseek · MIT |
1419 #120 of 416 |
Lab chat | Size unknown |
| 43 | qwen3.5-122b-a10b alibaba · Apache 2.0 |
1417 #123 of 416 |
Lab chat | 125.1 B~85 GB at 4-bit · 4 GPUs of 24 GB |
| 44 | deepseek-v3.1-terminus deepseek · MIT |
1417 #125 of 416 |
Lab chat | Size unknown |
| 45 | qwen3-next-80b-a3b-instruct alibaba · Apache 2.0 |
1417 #126 of 416 |
Lab chat | Size unknown |
| 46 | deepseek-v3.1-thinking deepseek · MIT |
1417 #128 of 416 |
Lab chat | Size unknown |
| 47 | qwen3-235b-a22b-thinking-2507 alibaba · Apache 2.0 |
1415 #131 of 416 |
Lab chat | Size unknown |
| 48 | kimi-k2-thinking-turbo moonshot · Modified MIT |
1414 #133 of 416 |
Lab chat | Size unknown |
| 49 | Inkling Small thinky · Apache 2.0 |
1413 #135 of 416 |
Free API · try it | Size unknown |
| 50 | mimo-v2-flash (non-thinking) xiaomi · MIT |
1411 #139 of 416 |
— | Size unknown |
| 51 | nvidia-nemotron-3.5-super-vl-120b-a12b-nvfp4 nvidia · OpenMDW-1.1 |
1409 #141 of 416 |
Lab chat | Size unknown |
| 52 | qwen3.5-27b alibaba · Apache 2.0 |
1409 #142 of 416 |
Lab chat | 27.8 B~20 GB at 4-bit · 1 GPU of 24 GB |
| 53 | minimax-m2.7 minimax · Modified MIT |
1405 #148 of 416 |
Lab chat | Size unknown |
| 54 | step-3.5-flash stepfun · Apache 2.0 |
1403 #150 of 416 |
— | Size unknown |
| 55 | hunyuan-hy3-preview tencent · tencent-hunyuan-community |
1402 #151 of 416 |
Lab chat | Size unknown |
| 56 | qwen3-vl-235b-a22b-thinking alibaba · Apache 2.0 |
1401 #152 of 416 |
Lab chat | 235.7 B~158 GB at 4-bit · 7 GPUs of 24 GB |
| 57 | qwen3.5-35b-a3b alibaba · Apache 2.0 |
1395 #161 of 416 |
Lab chat | Size unknown |
| 58 | mimo-v2-flash (thinking) xiaomi · MIT |
1395 #162 of 416 |
— | Size unknown |
| 59 | qwen3-235b-a22b-no-thinking alibaba · Apache 2.0 |
1393 #164 of 416 |
Lab chat | Size unknown |
| 60 | minimax-m2.1-preview minimax · MIT |
1392 #165 of 416 |
Lab chat | Size unknown |
Small models, no Arena votes yet
Official models from the labs' Hugging Face pages, 0.5 to 16 billion parameters. Arena has not voted on them, so they are ordered by downloads, which shows popularity, not quality. They run on 4 to 12 GB of GPU memory.
| Model | Downloads/month | Size | Run at home |
|---|---|---|---|
| Qwen3-0.6B Qwen · model |
30.0M | 0.8 B | ~3 GB at 4-bit · simulate |
| Qwen3-4B Qwen · model |
11.8M | 4 B | ~5 GB at 4-bit · simulate |
| Qwen3-8B Qwen · model |
9.1M | 8.2 B | ~7 GB at 4-bit · simulate |
| Qwen3-14B Qwen · model |
8.2M | 14.8 B | ~12 GB at 4-bit · simulate |
| Qwen3.5-4B Qwen · model |
7.9M | 4.7 B | ~5 GB at 4-bit · simulate |
| Qwen3.5-9B Qwen · model |
7.8M | 9.7 B | ~8 GB at 4-bit · simulate |
| gemma-3-1b-it google · model · license to accept |
1.8M | 1 B | ~3 GB at 4-bit · simulate |
| NVIDIA-Nemotron-3-Nano-4B-BF16 nvidia · model |
1.6M | 4 B | ~5 GB at 4-bit · simulate |
| SmolVLM2-500M-Video-Instruct HuggingFaceTB · model |
1.4M | 0.5 B | ~2 GB at 4-bit · simulate |
| MiniCPM5-2B openbmb · model |
1.3M | 2.5 B | ~4 GB at 4-bit · simulate |
| DeepSeek-R1-Distill-Qwen-1.5B deepseek-ai · model |
1.3M | 1.8 B | ~3 GB at 4-bit · simulate |
| gemma-3-4b-it google · model · license to accept |
1.1M | 4.3 B | ~5 GB at 4-bit · simulate |
| SmolLM3-3B HuggingFaceTB · model |
751K | 3.1 B | ~4 GB at 4-bit · simulate |
| granite-4.1-3b ibm-granite · model |
648K | 3.4 B | ~4 GB at 4-bit · simulate |
| Apertus-v1.5-8B swiss-ai · model · license to accept |
556K | 8.9 B | ~8 GB at 4-bit · simulate |
| Olmo-3-7B-Think allenai · model |
489K | 7.3 B | ~7 GB at 4-bit · simulate |
| DeepSeek-R1-0528-Qwen3-8B deepseek-ai · model |
465K | 8.2 B | ~7 GB at 4-bit · simulate |
| Apertus-8B-Instruct-2509 swiss-ai · model |
436K | 8.1 B | ~7 GB at 4-bit · simulate |
| Olmo-3-7B-Instruct allenai · model |
427K | 7.3 B | ~7 GB at 4-bit · simulate |
| DeepSeek-R1-Distill-Qwen-7B deepseek-ai · model |
385K | 7.6 B | ~7 GB at 4-bit · simulate |
57 of 144 models in Arena's Coding ranking are free to use. Votes as of 2026-10-16. Best free: Kimi K3 (Max), #14 overall.
| Free rank | Model | Arena score | Use free online | Run at home |
|---|---|---|---|---|
| 1 | Kimi K3 (Max) moonshot · Kimi K3 license |
1655 #14 of 144 |
Lab chat | 2779.9 B~1837 GB at 4-bit · server: 23 GPUs of 80 GB |
| 2 | MiMo-V2.6-Flash xiaomi · MIT |
1634 #17 of 144 |
— | 310.8 B~207 GB at 4-bit · 9 GPUs of 24 GB |
| 3 | Hy4 preview tencent · Apache 2.0 |
1631 #18 of 144 |
Lab chat | Size unknown |
| 4 | Qwen3.8-Flash-Next alibaba · qwen-community-1.0 |
1630 #19 of 144 |
Lab chat | 180 B~121 GB at 4-bit · 6 GPUs of 24 GB |
| 5 | MiMo-V2.6-Pro xiaomi · MIT |
1625 #20 of 144 |
— | 1024.2 B~678 GB at 4-bit · server: 9 GPUs of 80 GB |
| 6 | GLM-5.3 (Max) zai · MIT |
1621 #23 of 144 |
Lab chat | 753.3 B~499 GB at 4-bit · server: 7 GPUs of 80 GB |
| 7 | DeepSeek-V4.1-Flash (Max) deepseek · MIT |
1618 #24 of 144 |
Lab chat | 763.2 B~506 GB at 4-bit · server: 7 GPUs of 80 GB |
| 8 | GLM-5.3-Flash zai · MIT |
1605 #27 of 144 |
Lab chat | 321.3 B~214 GB at 4-bit · 9 GPUs of 24 GB |
| 9 | GLM-5.2 (Max) zai · MIT |
1605 #28 of 144 |
Lab chat | 753.3 B~499 GB at 4-bit · server: 7 GPUs of 80 GB |
| 10 | Qwen3.8-27B alibaba · Apache 2.0 |
1593 #29 of 144 |
Lab chat | 27.8 B~20 GB at 4-bit · 1 GPU of 24 GB |
| 11 | DeepSeek-V4-Pro (High, 20260813) deepseek · MIT |
1582 #33 of 144 |
Lab chat | 1598.8 B~1057 GB at 4-bit · server: 14 GPUs of 80 GB |
| 12 | DeepSeek-V4-Flash (High) deepseek · MIT |
1581 #34 of 144 |
Lab chat | 290.9 B~194 GB at 4-bit · 9 GPUs of 24 GB |
| 13 | Muse Spark Open (Max) meta · muse-model-license |
1511 #55 of 144 |
— | Size unknown |
| 14 | Kimi K2.6 moonshot · Modified MIT |
1510 #56 of 144 |
Lab chat | 1026.9 B~680 GB at 4-bit · server: 9 GPUs of 80 GB |
| 15 | GLM-5.1 zai · MIT |
1508 #57 of 144 |
Lab chat | 753.9 B~500 GB at 4-bit · server: 7 GPUs of 80 GB |
| 16 | Hy3 tencent · Apache 2.0 |
1506 #58 of 144 |
Lab chat | 298.8 B~199 GB at 4-bit · 9 GPUs of 24 GB |
| 17 | MiniMax-M3 minimax · MiniMax Community License |
1482 #64 of 144 |
Lab chat | 427 B~284 GB at 4-bit · server: 4 GPUs of 80 GB |
| 18 | MiMo-V2.5-Pro xiaomi · MIT |
1481 #65 of 144 |
— | 1023.2 B~677 GB at 4-bit · server: 9 GPUs of 80 GB |
| 19 | Kimi K2.7 Code moonshot · Modified MIT |
1474 #66 of 144 |
Lab chat | 1026.9 B~680 GB at 4-bit · server: 9 GPUs of 80 GB |
| 20 | DeepSeek-V4-Pro Preview (High) deepseek · MIT |
1464 #68 of 144 |
Lab chat | 1598.8 B~1057 GB at 4-bit · server: 14 GPUs of 80 GB |
| 21 | DeepSeek-V4-Pro deepseek · MIT |
1447 #72 of 144 |
Lab chat | 1598.8 B~1057 GB at 4-bit · server: 14 GPUs of 80 GB |
| 22 | MiMo-V2.5 xiaomi · MIT |
1438 #78 of 144 |
— | 310.8 B~207 GB at 4-bit · 9 GPUs of 24 GB |
| 23 | beam-1.0-501b-a23b reflection · Apache 2.0 |
1438 #79 of 144 |
— | Size unknown |
| 24 | Kimi K2.5 (Thinking) moonshot · Modified MIT |
1437 #80 of 144 |
Lab chat | 1026.9 B~680 GB at 4-bit · server: 9 GPUs of 80 GB |
| 25 | GLM-5 zai · MIT |
1435 #81 of 144 |
Lab chat | 753.9 B~500 GB at 4-bit · server: 7 GPUs of 80 GB |
| 26 | GLM-4.7 zai · MIT |
1435 #82 of 144 |
Lab chat | 358.3 B~238 GB at 4-bit · 10 GPUs of 24 GB |
| 27 | DeepSeek-V4-Flash Preview (High) deepseek · MIT |
1431 #84 of 144 |
Lab chat | 290.9 B~194 GB at 4-bit · 9 GPUs of 24 GB |
| 28 | Inkling thinky · Apache 2.0 |
1414 #87 of 144 |
Free API · try it | Size unknown |
| 29 | Inkling-Small thinky · Apache 2.0 |
1409 #88 of 144 |
Free API · try it | Size unknown |
| 30 | Kimi K2.5 (Instant) moonshot · Modified MIT |
1405 #90 of 144 |
Lab chat | 1026.9 B~680 GB at 4-bit · server: 9 GPUs of 80 GB |
| 31 | Qwen3.5-397B-A17B alibaba · Apache 2.0 |
1400 #93 of 144 |
Lab chat | 403.4 B~268 GB at 4-bit · server: 4 GPUs of 80 GB |
| 32 | MiniMax-M2.7 minimax · Modified MIT |
1399 #94 of 144 |
Lab chat | 228.7 B~153 GB at 4-bit · 7 GPUs of 24 GB |
| 33 | MiniMax-M2.5 minimax · Modified MIT |
1386 #99 of 144 |
Lab chat | 228.7 B~153 GB at 4-bit · 7 GPUs of 24 GB |
| 34 | MiniMax-M2.1 Preview minimax · MIT |
1384 #102 of 144 |
Lab chat | 228.7 B~153 GB at 4-bit · 7 GPUs of 24 GB |
| 35 | Gemma 4 31B google · Apache 2.0 |
1367 #106 of 144 |
Free API · try itLab chat | 31.3 B~23 GB at 4-bit · 1 GPU of 24 GB |
| 36 | DeepSeek-V3.2 (Thinking) deepseek · MIT |
1362 #107 of 144 |
Lab chat | 685.4 B~454 GB at 4-bit · server: 6 GPUs of 80 GB |
| 37 | Qwen3.5-122B-A10B alibaba · Apache 2.0 |
1361 #108 of 144 |
Lab chat | 125.1 B~85 GB at 4-bit · 4 GPUs of 24 GB |
| 38 | Gemma 4 26B A4B google · Apache 2.0 |
1359 #109 of 144 |
Free API · try itLab chat | 25.8 B~19 GB at 4-bit · 1 GPU of 24 GB |
| 39 | Qwen3.5-27B alibaba · Apache 2.0 |
1358 #110 of 144 |
Lab chat | 27.8 B~20 GB at 4-bit · 1 GPU of 24 GB |
| 40 | Hy3 preview tencent · tencent-hunyuan-community |
1357 #112 of 144 |
Lab chat | 298.8 B~199 GB at 4-bit · 9 GPUs of 24 GB |
| 41 | Muse Glimmer meta · Apache-2.0 |
1355 #113 of 144 |
— | Size unknown |
| 42 | Laguna M.1 poolside · Apache 2.0 |
1350 #114 of 144 |
— | Size unknown |
| 43 | GLM-4.6 zai · MIT |
1341 #116 of 144 |
Lab chat | 356.8 B~237 GB at 4-bit · 10 GPUs of 24 GB |
| 44 | MiMo-V2-Flash (Non-Thinking) xiaomi · MIT |
1330 #119 of 144 |
— | 309.8 B~206 GB at 4-bit · 9 GPUs of 24 GB |
| 45 | DeepSeek-V3.2 deepseek · MIT |
1326 #121 of 144 |
Lab chat | Size unknown |
| 46 | Kimi K2 Thinking Turbo moonshot · Modified MIT |
1322 #122 of 144 |
Lab chat | Size unknown |
| 47 | Laguna XS.2 poolside · Apache 2.0 |
1302 #123 of 144 |
— | Size unknown |
| 48 | MiniMax-M2 minimax · Apache 2.0 |
1297 #124 of 144 |
Lab chat | Size unknown |
| 49 | MiMo-V2-Flash (Thinking) xiaomi · MIT |
1293 #125 of 144 |
— | Size unknown |
| 50 | Qwen3-Coder-480B-A35B-Instruct alibaba · Apache 2.0 |
1275 #126 of 144 |
Lab chat | Size unknown |
| 51 | DeepSeek-V3.2-Exp deepseek · MIT |
1272 #127 of 144 |
Lab chat | Size unknown |
| 52 | Mistral Medium 3.5 mistral · Modified MIT |
1264 #128 of 144 |
— | Size unknown |
| 53 | Qwen3.5-35B-A3B alibaba · Apache 2.0 |
1254 #131 of 144 |
Lab chat | Size unknown |
| 54 | Trinity Large Thinking — · Apache 2.0 |
1238 #135 of 144 |
— | Size unknown |
| 55 | Mistral Large 3 mistral · Apache 2.0 |
1230 #136 of 144 |
— | Size unknown |
| 56 | Devstral 2 mistral · Modified MIT |
1197 #139 of 144 |
— | Size unknown |
| 57 | Granite 4.1 8B ibm · Apache 2.0 |
1193 #140 of 144 |
— | Size unknown |
49 of 165 models in Arena's Vision ranking are free to use. Votes as of 2026-10-09. Best free: glm-5.3-flash, #23 overall.
| Free rank | Model | Arena score | Use free online | Run at home |
|---|---|---|---|---|
| 1 | glm-5.3-flash zai · MIT |
1296 #23 of 165 |
Lab chat | 321.3 B~214 GB at 4-bit · 9 GPUs of 24 GB |
| 2 | kimi-k2.6 moonshot · Modified MIT |
1283 #33 of 165 |
Lab chat | 1026.9 B~680 GB at 4-bit · server: 9 GPUs of 80 GB |
| 3 | gemma-4-31b google · Apache 2.0 |
1277 #38 of 165 |
Free API · try itLab chat | 31.3 B~23 GB at 4-bit · 1 GPU of 24 GB |
| 4 | deepseek-v4.1-flash-max deepseek · MIT |
1277 #39 of 165 |
Lab chat | 763.2 B~506 GB at 4-bit · server: 7 GPUs of 80 GB |
| 5 | qwen3.8-27b alibaba · Apache 2.0 |
1271 #42 of 165 |
Lab chat | 27.8 B~20 GB at 4-bit · 1 GPU of 24 GB |
| 6 | kimi-k2.5-thinking moonshot · Modified MIT |
1269 #44 of 165 |
Lab chat | Size unknown |
| 7 | mimo-v2.6-pro xiaomi · MIT |
1264 #50 of 165 |
— | 1024.2 B~678 GB at 4-bit · server: 9 GPUs of 80 GB |
| 8 | qwen3.5-397b-a17b alibaba · Apache 2.0 |
1263 #54 of 165 |
Lab chat | 403.4 B~268 GB at 4-bit · server: 4 GPUs of 80 GB |
| 9 | gemma-4-26b-a4b google · Apache 2.0 |
1260 #55 of 165 |
Free API · try itLab chat | 25.8 B~19 GB at 4-bit · 1 GPU of 24 GB |
| 10 | mimo-v2.6-flash xiaomi · MIT |
1259 #56 of 165 |
— | 310.8 B~207 GB at 4-bit · 9 GPUs of 24 GB |
| 11 | kimi-k2.5-instant moonshot · Modified MIT |
1253 #59 of 165 |
Lab chat | Size unknown |
| 12 | minimax-m3 minimax · MiniMax Community License |
1253 #60 of 165 |
Lab chat | 427 B~284 GB at 4-bit · server: 4 GPUs of 80 GB |
| 13 | mimo-v2.5 xiaomi · MIT |
1247 #65 of 165 |
— | 310.8 B~207 GB at 4-bit · 9 GPUs of 24 GB |
| 14 | qwen3-vl-235b-a22b-instruct alibaba · Apache 2.0 |
1247 #66 of 165 |
Lab chat | 235.7 B~158 GB at 4-bit · 7 GPUs of 24 GB |
| 15 | qwen3.5-122b-a10b alibaba · Apache 2.0 |
1245 #70 of 165 |
Lab chat | 125.1 B~85 GB at 4-bit · 4 GPUs of 24 GB |
| 16 | qwen3.5-27b alibaba · Apache 2.0 |
1241 #72 of 165 |
Lab chat | 27.8 B~20 GB at 4-bit · 1 GPU of 24 GB |
| 17 | Inkling Small thinky · Apache 2.0 |
1235 #75 of 165 |
Free API · try it | Size unknown |
| 18 | mistral-medium-3.5 mistral · Modified MIT |
1223 #82 of 165 |
— | Size unknown |
| 19 | mistral-large-3 mistral · Apache 2.0 |
1221 #83 of 165 |
— | Size unknown |
| 20 | qwen3-vl-235b-a22b-thinking alibaba · Apache 2.0 |
1207 #90 of 165 |
Lab chat | 235.7 B~158 GB at 4-bit · 7 GPUs of 24 GB |
| 21 | step-3 stepfun · Apache 2.0 |
1176 #104 of 165 |
— | 321 B~214 GB at 4-bit · 9 GPUs of 24 GB |
| 22 | gemma-3-27b-it google · Gemma |
1164 #110 of 165 |
Lab chat | 27.4 B~20 GB at 4-bit · 1 GPU of 24 GB |
| 23 | glm-4.6v zai · MIT |
1164 #111 of 165 |
Lab chat | 107.7 B~73 GB at 4-bit · 4 GPUs of 24 GB |
| 24 | glm-4.5v zai · MIT |
1154 #116 of 165 |
Lab chat | 107.7 B~73 GB at 4-bit · 4 GPUs of 24 GB |
| 25 | qwen2.5-vl-32b-instruct alibaba · Apache 2.0 |
1152 #117 of 165 |
Lab chat | 33.5 B~24 GB at 4-bit · 1 GPU of 24 GB |
| 26 | mistral-small-2506 mistral · Apache 2.0 |
1142 #119 of 165 |
— | Size unknown |
| 27 | llama-4-maverick-17b-128e-instruct meta · Llama 4 |
1142 #120 of 165 |
— | 401.6 B~267 GB at 4-bit · server: 4 GPUs of 80 GB |
| 28 | mistral-small-3.1-24b-instruct-2503 mistral · Apache 2.0 |
1136 #123 of 165 |
— | Size unknown |
| 29 | llama-4-scout-17b-16e-instruct meta · Llama |
1118 #127 of 165 |
— | Size unknown |
| 30 | qwen2.5-vl-72b-instruct alibaba · Qwen |
1107 #128 of 165 |
Lab chat | Size unknown |
| 31 | pixtral-large-2411 mistral · MRL |
1089 #132 of 165 |
— | Size unknown |
| 32 | molmo-2-8b allenai · Apache 2.0 |
1081 #134 of 165 |
— | Size unknown |
| 33 | qwen2-vl-72b alibaba · Qwen |
1047 #139 of 165 |
Lab chat | Size unknown |
| 34 | molmo-72b-0924 allenai · Apache 2.0 |
1012 #143 of 165 |
— | Size unknown |
| 35 | pixtral-12b-2409 mistral · Apache 2.0 |
1009 #144 of 165 |
— | Size unknown |
| 36 | internvl2-26b — · MIT |
1005 #146 of 165 |
— | Size unknown |
| 37 | llama-3.2-vision-90b-instruct meta · Llama 3.2 |
1000 #147 of 165 |
— | Size unknown |
| 38 | c4ai-aya-vision-32b cohere · CC-BY-NC-4.0 |
996 #149 of 165 |
— | Size unknown |
| 39 | qwen2-vl-7b-instruct alibaba · Apache 2.0 |
990 #151 of 165 |
Lab chat | Size unknown |
| 40 | llama-3.2-vision-11b-instruct meta · Llama 3.2 |
966 #155 of 165 |
— | Size unknown |
| 41 | molmo-7b-d-0924 allenai · Apache 2.0 |
959 #156 of 165 |
— | Size unknown |
| 42 | internvl2-4b — · MIT |
941 #158 of 165 |
— | Size unknown |
| 43 | nvila-internal-15b-v1 nvidia · - |
940 #159 of 165 |
Lab chat | Size unknown |
| 44 | llava-v1.6-34b — · Apache 2.0 |
936 #160 of 165 |
— | Size unknown |
| 45 | cogvlm2-llama3-chat-19b zai · CogVLM2 |
928 #161 of 165 |
Lab chat | Size unknown |
| 46 | llava-onevision-qwen2-72b-ov — · Apache 2.0 |
923 #162 of 165 |
— | Size unknown |
| 47 | minicpm-v-2_6 — · Apache 2.0 |
912 #163 of 165 |
— | Size unknown |
| 48 | phi-3.5-vision-instruct microsoft · MIT |
851 #164 of 165 |
— | Size unknown |
| 49 | phi-3-vision-128k-instruct microsoft · MIT |
812 #165 of 165 |
— | Size unknown |
Small models, no Arena votes yet
Official models from the labs' Hugging Face pages, 0.5 to 16 billion parameters. Arena has not voted on them, so they are ordered by downloads, which shows popularity, not quality. They run on 4 to 12 GB of GPU memory.
| Model | Downloads/month | Size | Run at home |
|---|---|---|---|
| Qwen3.5-4B Qwen · model |
7.9M | 4.7 B | ~5 GB at 4-bit · simulate |
| Qwen3.5-9B Qwen · model |
7.8M | 9.7 B | ~8 GB at 4-bit · simulate |
| SmolVLM2-500M-Video-Instruct HuggingFaceTB · model |
1.4M | 0.5 B | ~2 GB at 4-bit · simulate |
| gemma-3-4b-it google · model · license to accept |
1.1M | 4.3 B | ~5 GB at 4-bit · simulate |
| Apertus-v1.5-8B swiss-ai · model · license to accept |
556K | 8.9 B | ~8 GB at 4-bit · simulate |
| gemma-3-12b-it google · model · license to accept |
365K | 12.2 B | ~10 GB at 4-bit · simulate |
| MiniCPM-V-4.6 openbmb · model |
204K | 1.3 B | ~3 GB at 4-bit · simulate |
| granite-vision-4.1-4b ibm-granite · model |
160K | 4 B | ~5 GB at 4-bit · simulate |
| GLM-4.1V-9B-Thinking zai-org · model |
125K | 10.3 B | ~9 GB at 4-bit · simulate |
| gemma-3n-E2B-it google · model · license to accept |
123K | 5.4 B | ~6 GB at 4-bit · simulate |
| SmolVLM2-2.2B-Instruct HuggingFaceTB · model |
122K | 2.2 B | ~3 GB at 4-bit · simulate |
| SmolVLM-500M-Instruct HuggingFaceTB · model |
111K | 0.5 B | ~2 GB at 4-bit · simulate |
| GLM-4.6V-Flash zai-org · model |
103K | 10.3 B | ~9 GB at 4-bit · simulate |
| MiniCPM-V-4_5 openbmb · model |
85K | 8.7 B | ~8 GB at 4-bit · simulate |
| Llama-3.1-Nemotron-Nano-VL-8B-V1 nvidia · model |
80K | 8.7 B | ~8 GB at 4-bit · simulate |
| MiniCPM-V-4.6-Thinking openbmb · model |
70K | 1.3 B | ~3 GB at 4-bit · simulate |
| Molmo2-8B allenai · model |
46K | 8.7 B | ~8 GB at 4-bit · simulate |
| North-Micro-Vision-Instruct CohereLabs · model |
46K | 2.5 B | ~4 GB at 4-bit · simulate |
| MiniCPM-V-4 openbmb · model |
41K | 4.1 B | ~5 GB at 4-bit · simulate |
| LFM2.5-VL-1.6B LiquidAI · model |
31K | 1.6 B | ~3 GB at 4-bit · simulate |
19 of 82 models in Arena's Image ranking are free to use. Votes as of 2026-10-07. Best free: qwen-image-2.1, #19 overall.
| Free rank | Model | Arena score | Use free online | Run at home |
|---|---|---|---|---|
| 1 | qwen-image-2.1 alibaba · qwen-research |
1223 #19 of 82 |
Lab chat | 7.1 B~7 GB at 4-bit · 1 GPU of 24 GB |
| 2 | ideogram-4.0-quality ideogram · Ideogram Open Model |
1205 #20 of 82 |
— | Size unknown |
| 3 | Cosmos3-Super-Text2Image (Agentic) nvidia · OpenMDW-1.1 |
1165 #27 of 82 |
Lab chat | Size unknown |
| 4 | Cosmos3-Super-Text2Image nvidia · OpenMDW-1.1 |
1152 #33 of 82 |
Lab chat | 64.6 B~45 GB at 4-bit · 2 GPUs of 24 GB |
| 5 | hunyuan-image-3.0 tencent · tencent-hunyuan-community |
1150 #35 of 82 |
Lab chat | 83 B~57 GB at 4-bit · 3 GPUs of 24 GB |
| 6 | flux-2-dev bfl · flux-non-commercial-license |
1146 #38 of 82 |
— | 32.2 B~23 GB at 4-bit · 1 GPU of 24 GB |
| 7 | qwen-image-2512 alibaba · Apache 2.0 |
1125 #44 of 82 |
Lab chat | 20.4 B~15 GB at 4-bit · 1 GPU of 24 GB |
| 8 | hidream-o1-image hidream · MIT |
1116 #47 of 82 |
— | 8.8 B~8 GB at 4-bit · 1 GPU of 24 GB |
| 9 | krea-2-turbo krea · krea-2-community-license |
1113 #52 of 82 |
— | 12.8 B~10 GB at 4-bit · 1 GPU of 24 GB |
| 10 | z-image-turbo alibaba · Apache 2.0 |
1083 #59 of 82 |
Lab chat | 6.2 B~6 GB at 4-bit · 1 GPU of 24 GB |
| 11 | flux-2-klein-9b bfl · flux-non-commercial-license |
1070 #62 of 82 |
— | 9.1 B~8 GB at 4-bit · 1 GPU of 24 GB |
| 12 | qwen-image-prompt-extend alibaba · Apache 2.0 |
1060 #63 of 82 |
Lab chat | Size unknown |
| 13 | qwen-image alibaba · Apache 2.0 |
1057 #66 of 82 |
Lab chat | 20.4 B~15 GB at 4-bit · 1 GPU of 24 GB |
| 14 | flux-2-klein-4b bfl · Apache 2.0 |
1030 #70 of 82 |
— | 3.9 B~5 GB at 4-bit · 1 GPU of 24 GB |
| 15 | glm-image zai · MIT |
1012 #76 of 82 |
Lab chat | 6.9 B~7 GB at 4-bit · 1 GPU of 24 GB |
| 16 | flux-1-dev-fp8 bfl · Open |
969 #78 of 82 |
— | Size unknown |
| 17 | flux-1-kontext-dev bfl · flux-1-dev-non-commercial-license |
941 #80 of 82 |
— | 11.9 B~10 GB at 4-bit · 1 GPU of 24 GB |
| 18 | stable-diffusion-v35-large — · Open |
938 #81 of 82 |
— | 8.1 B~7 GB at 4-bit · 1 GPU of 24 GB |
| 19 | bagel bytedance · Apache 2.0 |
898 #82 of 82 |
— | Size unknown |
7 of 48 models in Arena's Video ranking are free to use. Votes as of 2026-09-22. Best free: minimax-h3, #8 overall.
| Free rank | Model | Arena score | Use free online | Run at home |
|---|---|---|---|---|
| 1 | minimax-h3 minimax · minimax-h3-community-license-agreement |
1460 #8 of 48 |
Lab chat | 33.1 B~24 GB at 4-bit · 1 GPU of 24 GB |
| 2 | kandinsky-5.0-t2v-pro kandinsky · MIT |
1175 #37 of 48 |
— | 19.3 B~15 GB at 4-bit · 1 GPU of 24 GB |
| 3 | hunyuan-video-1.5 tencent · tencent-hunyuan-community |
1169 #38 of 48 |
Lab chat | 8.3 B~7 GB at 4-bit · 1 GPU of 24 GB |
| 4 | ltx-2-19b — · ltx-2-community-license-agreement |
1154 #41 of 48 |
— | 18.9 B~14 GB at 4-bit · 1 GPU of 24 GB |
| 5 | wan-v2.2-a14b alibaba · Apache 2.0 |
1132 #42 of 48 |
Lab chat | 27 B~20 GB at 4-bit · 1 GPU of 24 GB |
| 6 | kandinsky-5.0-t2v-lite kandinsky · MIT |
1113 #44 of 48 |
— | 2 B~3 GB at 4-bit · 1 GPU of 24 GB |
| 7 | mochi-v1 genmo · Apache 2.0 |
1007 #48 of 48 |
— | 10 B~9 GB at 4-bit · 1 GPU of 24 GB |
10 of 59 models in Arena's Image edit ranking are free to use. Votes as of 2026-10-06. Best free: qwen-image-2.1, #18 overall.
| Free rank | Model | Arena score | Use free online | Run at home |
|---|---|---|---|---|
| 1 | qwen-image-2.1 alibaba · qwen-research |
1369 #18 of 59 |
Lab chat | 7.1 B~7 GB at 4-bit · 1 GPU of 24 GB |
| 2 | hunyuan-image-3.0-instruct tencent · tencent-hunyuan-community |
1303 #26 of 59 |
Lab chat | 83 B~57 GB at 4-bit · 3 GPUs of 24 GB |
| 3 | qwen-image-edit alibaba · Apache 2.0 |
1242 #37 of 59 |
Lab chat | 20.4 B~15 GB at 4-bit · 1 GPU of 24 GB |
| 4 | qwen-image-edit-2511 alibaba · Apache 2.0 |
1234 #39 of 59 |
Lab chat | 20.4 B~15 GB at 4-bit · 1 GPU of 24 GB |
| 5 | flux-2-klein-9b bfl · flux-non-commercial-license |
1225 #42 of 59 |
— | 9.1 B~8 GB at 4-bit · 1 GPU of 24 GB |
| 6 | flux-2-dev bfl · flux-non-commercial-license |
1225 #43 of 59 |
— | 32.2 B~23 GB at 4-bit · 1 GPU of 24 GB |
| 7 | flux-2-klein-4b bfl · Apache 2.0 |
1187 #49 of 59 |
— | 3.9 B~5 GB at 4-bit · 1 GPU of 24 GB |
| 8 | flux-1-kontext-dev bfl · flux-1-dev-non-commercial-license |
1149 #53 of 59 |
— | 11.9 B~10 GB at 4-bit · 1 GPU of 24 GB |
| 9 | bagel bytedance · Apache 2.0 |
1027 #58 of 59 |
— | Size unknown |
| 10 | step1x-edit stepfun · Apache 2.0 |
998 #59 of 59 |
— | Size unknown |