Needle
Tiny on-device model for tool calls, extraction and speech input
What it is
Needle is a small foundation model, shipped as an 8-29 MB quantized binary, that picks and fills function calls, extracts typed fields from messy text, and returns sentence embeddings. A companion speech-to-text model, Whistle, runs in the same C++ engine so audio can go straight to tool calls. App and firmware developers building for phones, wearables, cars, robots and microcontrollers use it through a Python package or platform engines, and can fine-tune both models locally with LoRA.