← Back to Model Beat
Models·4d ago·all news from October 14, 2026

Fish Audio Releases Drama 3, a 35B Open-Weight MoE Text-to-Speech Model

Fish Audio has released Drama 3, a 35B mixture-of-experts model designed for text-to-speech tasks that utilizes 3B active parameters. The model supports 83 languages, accepts natural language instructions, and is available under a revenue-capped community license.

Covered by 2 sources

  • MMarkTechPost↗Asif Razzaq4d ago
  • AarXiv CS.AI↗Dawei Han, Shidong Li, Shijia Liao, Jiahua Liu, Songting Liu, Qingzheng Wang, Yuxuan Wang, Xin Chen, Yifan Cheng, Tianyu Li, Gangin Park, Xuehao Yang, Chang Zeng, Haoyang Zhang, Ruoyi Zhang, Xiangyu Zhang4d ago

Related stories

ModelsGemini 4 Argon: Built to solve your hardest problemsOct 13 · 11 sourcesModelsDeepseek V4.1 Flash 1017 beats larger rivals in agentic tasks while once again massively cutting per-task costsOct 16 · 14 sourcesModelsMeta Releases Open AI Model While Keeping Its Best Tech ClosedOct 15 · 9 sourcesModelsAlibaba Unveils $450 AI Glasses in Expanding Push Into GadgetsOct 13 · 21 sources