Third-Party Software & AI Models
Last updated: April 2026
InkWeave ships optional local AI features using open-source components and downloadable model weights. This page lists upstream projects, Hugging Face repositories used for downloads, and license references so you know where software comes from.
The exact builds, filenames, and default offerings may change with app updates; the in-app model list takes precedence. If you cannot reach Hugging Face from your region, a VPN or proxy may be required—see our Privacy Policy.
Inference engine (llama.cpp)
Local inference uses llama.cpp (including the llama-server HTTP server). Source code and license: github.com/ggml-org/llama.cpp
Downloadable GGUF models (Hugging Face)
The following models can be downloaded inside the app from Hugging Face. Each repository link points to the model card; individual GGUF files are fetched via Hugging Face’s CDN when you choose “Download”.
| Model (quantization) | Hugging Face repo | GGUF filename | License / terms (reference) |
|---|---|---|---|
| Qwen3 0.6B (Q8_0) | unsloth/Qwen3-0.6B-GGUF | Qwen3-0.6B-Q8_0.gguf | Qwen — Apache-2.0 (see model card) · Apache-2.0 |
| Qwen3 1.7B (Q4_K_M) | unsloth/Qwen3-1.7B-GGUF | Qwen3-1.7B-Q4_K_M.gguf | Qwen — Apache-2.0 (see model card) · Apache-2.0 |
| Qwen3 4B (Q4_K_M) | unsloth/Qwen3-4B-GGUF | Qwen3-4B-Q4_K_M.gguf | Qwen — Apache-2.0 (see model card) · Apache-2.0 |
| Qwen3 8B (Q4_K_M) | unsloth/Qwen3-8B-GGUF | Qwen3-8B-Q4_K_M.gguf | Qwen — Apache-2.0 (see model card) · Apache-2.0 |
| Gemma 3 4B (Q4_K_M) | unsloth/gemma-3-4b-it-GGUF | gemma-3-4b-it-Q4_K_M.gguf | Gemma — Google Gemma Terms of Use · Gemma Terms |
Trademarks
Qwen and Gemma are trademarks of their respective owners. Unsloth provides community GGUF builds; InkWeave is not affiliated with Hugging Face, Unsloth, Alibaba, or Google.