# Gradium releases Phonon 31M, a tiny on‑device TTS model

> The English Phonon 31m model (31 M parameters, 64.5 MB) runs in ~7 ms on a laptop CPU and is now Apache 2.0‑licensed.

Oossa · 2026-10-22 · https://oossa.com/en/gradium-releases-phonon-31m-a-tiny-on-device-tts-model

Gradium has open‑sourced its Phonon family of text‑to‑speech models. The English Phonon 31m model contains 31 million parameters, fits in a 64.5 MB file, and can start speaking in about 7 ms on a typical laptop CPU. The code and model weights are released under the permissive Apache 2.0 license, allowing anyone to use or modify them.

## The facts

- 31 M parameters; file size 64.5 MB
- First spoken output in ~7 ms on a laptop CPU

## Why it matters

Developers can now add fast, offline voice synthesis to apps without paying for cloud services.

## Sources & references

1. [Gradium Open-Sources Phonon, a 31M-Parameter On-Device TTS Model Under Apache 2.0](https://www.marktechpost.com/2026/10/22/gradium-phonon-31m-on-device-tts/) – MarkTechPost, 2026-10-22

Last updated: 2026-10-22