Whistle: Speech to Text in 16.9 MB

A 16.9 MB open speech model runs on CPU in the same engine as Needle, so one binary can go from audio to tool calls on phones, wearables and microcontrollers without cloud calls.

• embedding — A list of numbers representing a piece of text's meaning, so that similar meanings end up numerically close and can be searched.

• articleStrands Decider 2B: a small, open-source, decision model
• articleEmber-1

• articleDesert Ant Labs: local, fast models that run on devicewillwhitedc
• articleCactus Needle 3: 8-29MB automation models can match DeepSeek V4 FlashHenryNdubuaku
• repoPlainsay – 12 MB Mac dictation with local Whisper and Parakeetconrader

Checking sign-in…

Loading comments…