Announcement
Introducing Phonon-1
An open speech recognition model for English: a 415 MB download that runs on a laptop or a datacenter GPU and transcribes an hour of audio in about two and a half minutes.
Speech recognition · Available nowPhonon-1
An open speech recognition model for English: a 415 MB download that runs on a laptop or a datacenter GPU and transcribes an hour of audio in about two and a half minutes.
- 415 MB
- Download, 455 MB on disk
- 2.640 %
- LibriSpeech test-clean
- 23.9×
- Realtime, median of nine corpora
- Apache-2.0
- Open weights
Phonon-1 is an open speech recognition model for English. It downloads in 415 MB, runs on a laptop or a datacenter GPU, and transcribes an hour of audio in about two and a half minutes. It was trained at 2.4 bits per weight from the start, and it is the second model in the lab’s low-bit lane after Neutrino-1.
Output arrives punctuated and capitalized from the model itself. Phonon-1 powers Detta, the Fermion Research dictation app for the Mac.
- Phonon-1the default model
- 415 MB download · 455 MB on disk
- Phonon-1 Microthe smallest model; the accuracy trade is in the table below
- 285 MB download · 331 MB on disk
- Language
- English
- Audio inputmicrophone or file
- 16 kHz
- Output
- Punctuated, capitalized text
Benchmarks
| Benchmark | Phonon-1415 MB | Phonon-1 Micro285 MB | Parakeet-0.6B 4-bit637 MB | Moonshine base248 MB | Whisper large-v3-turbo1,619 MB | Whisper small967 MB | wav2vec2-large1,262 MB | Qwen3-ASR teacher1,569 MB |
|---|---|---|---|---|---|---|---|---|
| LibriSpeech test-clean | 2.640 | 3.002 | 2.186 | 3.417 | 2.10 | 3.4† | 2.8† | 2.235 |
| LibriSpeech test-other | 5.699 | 6.511 | 3.937 | 8.262 | 4.07 | 7.6† | 6.3† | 4.618 |
| TED-LIUM | 3.421 | 3.878 | 2.829 | 5.272 | – | – | – | 2.889 |
| SPGISpeech | 4.163 | 4.858 | 4.104 | 5.731 | 2.79† | – | 13.31† | 3.074 |
| VoxPopuli | 8.394 | 9.177 | 6.345 | 10.470 | 11.22† | – | – | 7.151 |
| GigaSpeech | 11.396 | 11.882 | 9.614 | 12.114 | 8.52† | – | – | 9.321 |
| Earnings-22 | 12.571 | 14.771 | 11.190 | 17.872 | 11.07† | – | 36.28† | 11.188 |
| AMI | 13.084 | 14.094 | 12.723 | 17.790 | 15.16† | – | – | 12.560 |
| Macro, eight benchmarks | 7.67 | 8.52 | 6.62 | 10.1 | – | – | – | 6.63 |
Speed
Median 23.9× realtime across nine corpora on a base M5 MacBook Air.
- Phonon-1
- p50 262 ms · p95 629 ms
- Fastest build of the family
- p50 203 ms · p95 446 ms
Run it
$ pip install fermion-research$ fermion transcribe recording.wav
One-shot file transcription. fermion listen transcribes the microphone live; fermion serve exposes an OpenAI-compatible transcription route.
$ fermion listen$ fermion serve
$ docker run --rm --gpus all ghcr.io/fermionresearch/phonon-cuda:latest \transcribe recording.wav
The container transcribes recordings of any length, streams over WebSocket, and serves the same endpoint as fermion serve.
The same weights run on a Mac (via MLX) or an NVIDIA GPU; the CUDA runtime and Docker image are in the GitHub repository.
Weights on Hugging Facefermion-research on PyPICUDA runtime on GitHub
Availability
Phonon-1 ships as public weights on Hugging Face, a command line on PyPI, and the CUDA runtime on GitHub. No waitlist, no gated preview.
Speech recognition that stays on the device.
Open weightsApache-2.0
License
Open weights under the Apache License 2.0. Commercial use, modification, fine-tuning, and redistribution are permitted, with no access request and no acceptance form. The model is a derivative of Qwen/Qwen3-ASR-0.6B, itself Apache-2.0; the command line is Apache-2.0 too.
Citation
@misc{fermionresearch2026phonon,
title = {Phonon-1},
author = {{Fermion Research}},
year = {2026},
url = {https://fermionresearch.com/models/phonon-1/}
}