• When you select a capability, models compatible with that task are listed.
• If a model isn't on your device, a "Download" button will be visible after you expand the model item. Tapping it will initiate the download.
• Once downloaded, the button will change to "Try It”, allowing you to use the model.
• Feel free to switch apps or lock your device. The download will continue in the background. We'll send a notification when it's done.

• Some models (e.g., Gemma 3) are gated and require you to log in with your Hugging Face (HF) account and agree to their specific license terms before downloading.
• The app will prompt you to log in via a web browser flow when you attempt to download such a model.
• Successfully logging in and accepting terms will allow the download to proceed. Your HF token is stored securely for future downloads.

For LLM tasks, after the model generates a response, you can view performance metrics:

• Look for a "Stats" button below the response.
• Metrics typically include:
  • Time to First Token (TTFT)
  • Prefill Speed (tokens/s)
  • Decode Speed (tokens/s)
  • Latency (sec)

• To delete a downloaded model, simply click on the delete icon next to the model in the model selection screen.