Draggy
Models
Draggy offers these model families by default. Each size is a separate model on Hugging Face. The license column shows what each model's page listed in October 2026, and the model's page is the one that counts.
Picked for your hardware
During setup, Draggy picks one of these for your computer, based on its graphics card and memory. The models page in the wiki explains how.
| Family | Maker | Sizes offered | License |
|---|---|---|---|
| Qwen 3.5 | Alibaba (Qwen) | 0.8B, 2B, 4B, 9B, 122B MoE | Apache 2.0 |
| Gemma 4 | E2B, E4B, 12B, 31B | Apache 2.0 | |
| Ministral 3 | Mistral AI | 14B | Apache 2.0 |
| GPT-OSS | OpenAI | 20B MoE, 120B MoE | Apache 2.0 |
| Qwen 3.6 | Alibaba (Qwen) | 35B MoE (Q4, Q8) | Apache 2.0 |
| Qwen 3.8 | Alibaba (Qwen) | 27B (Q4, Q8) | Apache 2.0 |
MoE means mixture of experts: the model is large in total, but uses only part of its parameters for each token.
Listed when you search
Search lists these families when your search matches their name or description. Other GGUF models on Hugging Face appear in the same results.
| Family | Maker | Sizes offered | License |
|---|---|---|---|
| Llama 3.1 | Meta | 8B, 70B | Llama 3.1 Community License |
| Llama 3.2 | Meta | 1B, 3B | Llama 3.2 Community License |
| DeepSeek-R1 | DeepSeek | 1.5B, 7B, 8B, 14B, 32B, 70B | See the model page |
| Qwen 2.5 | Alibaba (Qwen) | 0.5B, 1.5B, 3B, 7B, 14B, 32B | Apache 2.0 (3B: Qwen Research License) |
| Phi-4 | Microsoft | 14B | MIT |
| Mistral | Mistral AI | 7B (v0.3) | Apache 2.0 |
| Gemma 2 | 2B, 9B, 27B | Gemma Terms of Use | |
| nomic-embed-text | Nomic AI | v1.5, embedding model | Apache 2.0 |
Any GGUF model
Any GGUF model on Hugging Face can be added. The model list shows the top 100 trending GGUF models, and search finds others by name. The llama engine reads GGUF files only.
Draft models
When you download some models, Draggy also fetches a small draft model from the same family to speed up replies. These are Qwen3 0.6B, Qwen3.5 0.8B, Qwen2.5 0.5B, Llama 3.2 1B and Gemma 3 270M.