Tech News
Whistle: What Is Cactus’s 16.9 MB Voice Model
Cactus introduced the Whistle model, a 16.9 MB speech recognition system that runs on CPU without an internet connection. It transcribes seven languages, marks per-word timestamps, and shares the C++ engine with Cactus’s Needle text model. On an Apple M4 Pro it decodes at 1,319 tokens per second, ahead of Whisper base and Moonshine tiny v2.









