Local models.
Same picker.
Runs on your hardware. Lives in pi.
SwitchboardLocal is a pi extension. Paste a Hugging Face link, the model downloads, and it appears in /model at 0x. When you need a frontier model, pick one from the same list.
Local inference, priced at 0x.
Why run it local?
Local models in the same picker
Paste a Hugging Face link. The model downloads, then appears in /model next to your hosted catalog, at $0. No account and no keys.
Zero metering, zero fee
Local inference runs on your hardware, so there is nothing to meter and no routing fee. Unlimited calls, priced at 0x.
Nothing leaves the machine
Prompts and completions stay on-device. The only network traffic is the download from Hugging Face. After that, no provider, no hop, no third party.
Fall back without leaving pi
When a request needs a frontier model, pick one from the same list. Local first, hosted when it counts, same session.
Install
One picker.
Local or hosted.
Install the extension, paste a Hugging Face link, and the model shows up in /model. MLX on Apple Silicon, GGUF on macOS, Windows, and Linux. Hosted Switchboard models stay in the same list.
View the extension →pi install https://github.com/Valni-Labs/switchboard-local/switchboard-local huggingface.co/mlx-community/Llama-3.2-1B-Instruct-4bitStart with SwitchboardLocal
Install the pi extension, paste a Hugging Face link, run. Local models cost 0x.