Local Models

Agora can import GGUF language models for on-device llama.cpp inference.

Import

Open Settings → Providers → Local and import a GGUF file. Agora copies the model into app-managed storage; an optional multimodal projection file can be associated when supported. Imported models are enabled automatically and can be selected from Settings → Models or the chat model picker.

Configure and use

Model-specific context and generation capability depend on the GGUF and available device memory. Large context sizes and models require more RAM and may be slow or fail on constrained devices.

Delete

Deleting an imported local model removes Agora's managed GGUF and associated projection file. It does not delete an unrelated source file outside Agora's managed storage.

Review Models, Generation, and Privacy & Security. The separate Alpine sandbox is F-Droid-only, but it is not required for llama.cpp model inference.