Artificial intelligence / news brief
Gradium reports 200,000 downloads for its on-device Phonon TTS models
- Event
- Published

Gradium says its Phonon text-to-speech models passed 200,000 downloads in two weeks and now lead Hugging Face's CPU streaming benchmark for time to first audio. The 31M English checkpoint is Apache 2.0, while the multilingual 40M variants remain non-commercial; the models originally launched October 20, so this is an adoption and evaluation update rather than a new release.
Why it matters
Small offline speech models can reduce cloud latency and keep text local, but automated leaderboard scores do not establish listener preference or real-device performance.
Sources
- A Closer Look at Phonon TTS (external site)Gradium · primary source ·
- Gradium’s On-Device Phonon TTS Passes 200,000 Downloads in Two Weeks (external site)Unite.AI · coverage ·