Knowledge Share
Technical articles, tutorials, and insights
One Open Source Project a Day (No. 202): Needle 2 — A 14 MB On-Device Tool-Calling Model
Needle 2 is a 45M-parameter tool-calling model from Cactus Compute, packaged as a single 14 MB binary that runs in a fixed 28 MB of session RAM. It matches FunctionGemma 270M on tool-calling benchmarks at one-sixth the size, and runs at 300+ tok/s on sub-$200 Android phones.
Open Source Project of the Day (Part 11): Supertonic - Lightning-Fast On-Device Multilingual TTS, Running Speech Synthesis Locally at Full Speed
A deep dive into Supertonic, an open-source on-device TTS system with 2.6k+ Stars — achieving blazing-fast speech synthesis via ONNX Runtime, supporting 5 languages, with built-in text normalization, no cloud required, running entirely on-device
Open Source Project of the Day (Part 9): NexaSDK - Cross-Platform On-Device AI Runtime for Running Frontier Models Locally
A deep dive into NexaSDK, a cross-platform on-device AI runtime supporting GPU, NPU, and CPU across PC, mobile, and IoT devices — running frontier LLM and VLM models locally with high efficiency