# Third-Party Notices

The licenses and attributions in this document apply only to bundled third-party software and bundled third-party model artifacts. They do not license Untethered-LLM itself. Review `App License` in Settings for the app's licensing boundary.

## Included Runtime Dependencies

### llama.swift

- Project: `https://github.com/mattt/llama.swift`
- Version: `2.8252.0`
- License: MIT
- Use: Swift integration layer for local GGUF inference on Apple platforms.

### llama.cpp

- Project: `https://github.com/ggml-org/llama.cpp`
- Distributed via: `llama.swift` binary artifact
- License: MIT
- Use: Core on-device inference runtime.

### ZIPFoundation

- Project: `https://github.com/weichsel/ZIPFoundation`
- Version: `0.9.20`
- License: MIT
- Use: DOCX archive extraction for local document processing.

## Bundled Model Attribution

### Huihui-Qwen3-4B-Instruct-2507-abliterated-Q4_K_M

- Bundled file: `Huihui-Qwen3-4B-Instruct-2507-abliterated.Q4_K_M.gguf`
- Upstream GGUF source: `mradermacher/Huihui-Qwen3-4B-Instruct-2507-abliterated-GGUF`
- Base model family: `Qwen/Qwen3-4B`
- License: Apache License 2.0
- Use: Default bundled higher-capability model for fully offline chat.

### Qwen3-4B-Q5_K_M

- Bundled file: `Qwen3-4B-Q5_K_M.gguf`
- Upstream GGUF source: `Qwen/Qwen3-4B-GGUF`
- Base model family: `Qwen/Qwen3-4B`
- License: Apache License 2.0
- Use: Second bundled higher-capability model for fully offline chat.

### SmolLM2-135M-Instruct-Q2_K

- Bundled file: `SmolLM2-135M-Instruct-Q2_K.gguf`
- Upstream GGUF source: `unsloth/SmolLM2-135M-Instruct-GGUF`
- Base model family: `HuggingFaceTB/SmolLM2-135M-Instruct`
- License: Apache License 2.0
- Use: Bundled fallback model for lower-memory devices and recovery scenarios.

## Full License Texts

The full MIT notices and Apache License 2.0 text are bundled separately in Settings so recipients can review the exact third-party license terms shipped with the app. Those texts apply to the listed third-party components and model artifacts only.