Ant Group's Ming-3.0-tts tops open-weight TTS leaderboards with 93.2% pronunciation robustness
ArtificialAnlys · x · 2026-10-20
In Artificial Analysis' latest TTS evaluations, Ant Group's Ming-3.0-tts becomes the leading open-weights TTS model:
- Specs: a dense 0.5B model with preset voices and stateless voice cloning, plus a 7.9B MoE variant (Ming-3.0-tts-tiny); weights on Hugging Face under MIT.
- Pronunciation Robustness: #1 of 31 models at 93.2%, tied with Fish Audio Drama 3 Preview and ahead of Eleven v4 (91.7%); ranks #1 in Expanding Shorthand (96.5%) and Preserving Exact Sequences (90.6%).
- Controlled Voice Arena: #1 open weights with Elo 1,069, ahead of Drama 3 Preview (1,059) and Breeze TTS 2 (1,019); #13 overall.
- Provider Voice Arena: Elo 1,188, #3 among open weights and #19 overall, strongest in customer service and assistant prompts.
Ming-3.0-tts takes #1 open-weights spots on two of three TTS leaderboards.
Related event: Ant's Ming-3.0-tts Tops Open-Weight TTS Leaderboards(2 posts)→
More from Multimodal
- Fréchet Forcing trains long-horizon world models from short clips — Yian Zhao · 2026-10-20
- VideoGen's Model 1 claims 10-minute multi-shot video from one prompt, at half the cost — lmoroney · 2026-10-20
- Building 3D worlds with Gemini 4 Argon and Blender is my new favourite hobby — jacalulu · 2026-10-20
- Zelda meets Pikachu in a slick AI-generated video powered by Breeze TTS2 and Krea 2 — dramaton42 · 2026-10-20
- AI-generated boxing sparring video fools viewers with realism — Electrical-Theme2847 · 2026-10-20
- One-shot ASCII poster generated in a single Opus prompt — zsakib_ · 2026-10-20