Ant Group's Ming-3.0-tts Tops Open Weights TTS Leaderboards
A 0.5B model beating much larger commercial systems on pronunciation suggests small, MIT-licensed TTS can now be self-hosted for production voice applications at minimal compute cost. Developers building voice agents or dubbing pipelines in the region can download both variants from Hugging Face and run them locally, avoiding per-character API fees from providers like ElevenLabs. The main trade-off is that closed models still lead overall, with Ming-3.0-tts only #13 on Controlled Voice.
Ant Group released Ming-3.0-tts, an open weights text-to-speech model, in two variants: a dense 0.5B parameter model (Ming-3.0-tts-0.5B) and a 7.9B mixture-of-experts variant (Ming-3.0-tts-tiny). On the Artificial Analysis Controlled Voice TTS Arena, the 0.5B model is the top open weights entry with an Elo of 1,069 across 1,827 appearances, ahead of Fish Audio's Drama 3 Preview at 1,059 and Breeze TTS 2 at 1,019. It also ranks #1 of 31 models on the Pronunciation Robustness leaderboard with 93.2%, level with Drama 3 Preview and ahead of Eleven v4 at 91.7%. On the Provider Voice leaderboard it ranks #3 among open weights models with an Elo of 1,188. Weights are on Hugging Face under the MIT License, with preset voices and stateless voice cloning support.
Ant Group's Ming-3.0-tts is the new #1 open weights model in the Artificial Analysis Controlled Voice TTS Arena and takes the #1 spot on our Pronunciation Robustness leaderboard, level with Fish Audio's Drama 3 Preview and ahead of Eleven v4
Ming-3.0-tts is the new open weights Text to Speech model from @AntGroup team, a dense 0.5B parameter model with preset voices and stateless voice cloning. Weights are available on Hugging Face under the MIT License: Ming-3.0-tts-0.5B, alongside a 7.9B mixture-of-experts variant, Ming-3.0-tts-tiny.
Key takeaways:
➤ Pronunciation Robustness: Ming-3.0-tts scores 93.2%, ranking #1 of 31 models, level with Fish Audio Drama 3 Preview at 93.2% and ahead of Eleven v4 at 91.7%. It ranks #1 in Expanding Shorthand and Preserving Exact Sequences.
➤ Controlled Voice: Ming-3.0-tts is the #1 open weights model with an Elo of 1,069 (+16/-16) across 1,827 appearances, ahead of Drama 3 Preview at 1,059 and Breeze TTS 2 at 1,019. It ranks #13 overall.
➤ Provider Voice: Ming-3.0-tts has an Elo of 1,188 (+15/-15) across 1,957 appearances, ranking #3 among open weights models behind Drama 3 Preview at 1,245 and Breeze TTS 2 at 1,220, and #19 overall.
➤ Open weights: Ming-3.0-tts is the #1 open weights model on two of our three Text to Speech leaderboards, Controlled Voice and Pronunciation Robustness, and #3 on Provider Voice.
See more details and listen to samples below 🧵
