Moonshine
Tiny speech-recognition models — 27M parameters, built for microcontrollers and cheap hardware.
About
Moonshine is a family of tiny automatic speech recognition models from Useful Sensors, released in October 2024. The smallest variant has just 27 million parameters, yet the team reports accuracy competitive with much larger models on English transcription. It is designed for real-time transcription on severely constrained hardware, and the weights are MIT licensed. The tradeoff is English-only coverage and accuracy that trails large models on noisy or accented speech.
Specs
Context window
Pricing
Free quota
API access
Weights
Released
Benchmarks
Team-reported: moonshine-tiny (27M) beats Whisper tiny on Fleurs and Common Voice WER at a fraction of the size.
Provider-reported unless a source says otherwise. Benchmarks are marketing-adjacent by nature — treat them as directional.
Use cases
Trust score — 7/10
Scored 7/10 (verified tier, metrics-v1). Evidence: Released October 2024 per HF model card; MIT license per third-party tables — verify on card; English-only; tiny 27M and base 61M variants. License: MIT (per third-party model tables; confirm on the HF card)..
How the trust score is built
Full methodology →- Adoption · stars/downloads, log-scaled vs peers
- 6.0/10
- Maintenance · recency, releases, issue load
- 6.0/10
- Openness · license + docs
- 9.0/10
- Manual review · sanity read + red-flag check
- 6.0/10
- · Released October 2024 per HF model card
- · MIT license per third-party tables — verify on card
- · English-only; tiny 27M and base 61M variants