Tiny ASR Model Whistle Runs in Browsers at 16.9MB

Cactus-Compute's on-device ASR model Whistle trended on Hugging Face. With 55M parameters and 2-bit quantization, it is only 16.9MB—1/9 the size of Whisper base—while exceeding its accuracy and running in browsers via WASM.

2026-10-05 ~ 2026-10-06 · 2 related posts

• Whistle, an on-device speech recognition model, trends on Hugging Face — Cactus-Compute · 2026-10-05
• Whistle: a 16.9MB speech-to-text model that beats Whisper base at 9x smaller size and 6x speed — Henrie_the_dreamer · 2026-10-06