Best permissively licensed models
Open weights and open licence are different things. These carry licences that permit commercial use without a bespoke agreement.
From the file· live filter over real dataFrom the file· 40 models
How this is ranked
Filtered to SPDX licences that permit commercial use, then ranked by downloads. Always read the licence yourself — this is a filter, not legal advice.
Best permissively licensed models
| # | Model | Params○ | Smallest quant● | Licence○ |
|---|---|---|---|---|
| 1 | Qwen3-Coder-30B-A3B-InstructMoE | 30.5B | 7.46 GiB | apache-2.0 |
| 2 | Qwen3.6-27B | 27.8B | 8.74 GiB | apache-2.0 |
| 3 | Qwen3.6-35B-A3BMoE | 36.0B | 8.77 GiB | apache-2.0 |
| 4 | Qwen3.6-27B-Fable-Fusion-711-Uncensored-Heretic-NM-DAU-MTP | 27.8B | 10.12 GiB | apache-2.0 |
| 5 | Qwen3.8-27B | 27.8B | 8.39 GiB | apache-2.0 |
| 6 | Qwen3.5-9B | 9.7B | 2.97 GiB | apache-2.0 |
| 7 | gemma-4-26B-A4B-itMoE | 26.5B | 8.99 GiB | apache-2.0 |
| 8 | gemma-4-12B-it | 12.0B | 3.92 GiB | apache-2.0 |
| 9 | DeepSeek-V4-FlashMoE | 291B | 76.87 GiB | mit |
| 10 | Qwen3.5-4B | 4.7B | 1.42 GiB | apache-2.0 |
| 11 | parakeet-unified-en-0.6b | 618M | 0.44 GiB | cc-by-4.0 |
| 12 | gemma-4-12B-it-qat-q4_0-unquantized | 12.0B | 6.50 GiB | apache-2.0 |
| 13 | Qwythos-9B-Claude-Mythos-5-1M | 9.4B | 5.38 GiB | apache-2.0 |
| 14 | gemma-4-E4B-it | 8.0B | 3.30 GiB | apache-2.0 |
| 15 | Qwen3-30B-A3B-Thinking-2507MoE | 30.5B | 7.05 GiB | apache-2.0 |
| 16 | gemma-4-31B-it | 31.3B | 7.95 GiB | apache-2.0 |
| 17 | Muse-Glimmer-30B | 29.8B | 8.31 GiB | apache-2.0 |
| 18 | Qwen3-8B | 8.2B | 2.12 GiB | apache-2.0 |
| 19 | gemma-4-26B-A4B-it-qat-q4_0-unquantizedMoE | 26.5B | 13.45 GiB | apache-2.0 |
| 20 | cohere-transcribe-03-2026 | 2.1B | 1.41 GiB | apache-2.0 |
| 21 | Qwen3.5-122B-A10BMoE | 125B | 26.92 GiB | apache-2.0 |
| 22 | Qwen3.6-35B-A3B-Uncensored-HauhauCS-Aggressive | 34.7B | 34.37 GiB | apache-2.0 |
| 23 | DeepSeek-V4-Flash-0731MoE | 304B | 76.87 GiB | mit |
| 24 | Qwen3.5-0.8B | 873M | 0.31 GiB | apache-2.0 |
| 25 | Qwen-AgentWorld-35B-A3BMoE | 34.7B | 10.71 GiB | apache-2.0 |
| 26 | gemma-4-E2B-it | 5.1B | 2.13 GiB | apache-2.0 |
| 27 | gemma-4-31B-it-qat-q4_0-unquantized | 32.7B | 16.44 GiB | apache-2.0 |
| 28 | gpt-oss-20bMoE | 21.5B | 10.68 GiB | apache-2.0 |
| 29 | KAT-Coder-V2.5-DevMoE | 34.7B | 9.11 GiB | apache-2.0 |
| 30 | gemma-4-E4B-it-qat-q4_0-unquantized | 7.9B | 4.80 GiB | apache-2.0 |
| 31 | Qwen3-VL-30B-A3B-InstructMoE | 31.1B | 7.60 GiB | apache-2.0 |
| 32 | Qwen3-30B-A3BMoE | 30.5B | 7.59 GiB | apache-2.0 |
| 33 | Qwopus3.6-35B-A3B-v1MoE | 36.0B | 6.97 GiB | apache-2.0 |
| 34 | Qwen3.5-9B-The-Defiant-Fable-Uncensored-Heretic-NEO-IMATRIX-MAX-MTP | 9.7B | 9.21 GiB | apache-2.0 |
| 35 | parakeet-tdt-0.6b-v3 | 627M | 0.39 GiB | cc-by-4.0 |
| 36 | Qwen3.5-35B-A3BMoE | 36.0B | 8.77 GiB | apache-2.0 |
| 37 | Qwen3-VL-8B-Instruct-abliterated-v1 | 8.8B | 1.97 GiB | apache-2.0 |
| 38 | ced-base | 86M | 0.12 GiB | apache-2.0 |
| 39 | Hy3MoE | 299B | 59.65 GiB | apache-2.0 |
| 40 | Qwen2.5-7B-Instruct | 7.6B | 2.59 GiB | apache-2.0 |
Best local models for codingBest local reasoning modelsBest models for an 8GB cardBest models for long contextBest models under 4GBLargest models with a runnable quantizationBest local speech recognition modelsBest local text-to-speech modelsBest local vision-language modelsBest local image generation models