AI News
Zyphra Open-Sources ZONOS3-Mini TTS Model
November 13, 2026 AI Weekly
ModelsTools
Key Takeaways
- 1Zyphra released ZONOS3-Mini, a 100M-parameter Mixture-of-Experts text-to-speech model.
- 2The model achieves an impressive 80ms time-to-first-audio latency.
- 3It currently leads the ZTTS2-Eval benchmark for open-weight speech synthesis models.