Prism – AI Playground

Prism is a native AI workspace for running models locally on iPhone and iPad.

Chat with Apple Intelligence on supported devices, or run downloaded and imported GGUF models through llama.cpp. Prism keeps conversations, model files, and settings on your device, without requiring an account, subscription, or cloud inference service.

Use Prism to:

- Chat with local models in a polished native interface
- Choose from a curated model library
- Download models in the background
- Import GGUF files from Files or direct URLs
- Reopen and search local conversation history
- Export chats as Markdown or JSON
- Inspect prompts in Developer Mode
- Tune inference settings like context, temperature, top-p, top-k, repeat penalty, and max tokens
- Track generation stats and per-model performance

Prism is built for developers, local AI enthusiasts, prompt tinkerers, and privacy conscious users who want control over the model running on their device.

Once a supported model is available locally, Prism is designed around on-device use. Internet access is only needed for model acquisition flows such as downloading from a curated source or direct model URL.