# inclusionAI releases open weights for five Ming-3.0 speech models



Ant Group



Ant Group’s inclusionAI released five Ming-3.0 models for speech synthesis, audio understanding and speech recognition. Weights are available on Hugging Face and ModelScope.


## Details

The release includes five models across three families. For speech synthesis, a 0.5B model supports voice design and paralinguistic cues, while the Tiny model has 7.9B total and 1.3B active parameters for long-form speech. For audio understanding, Flash has 126B total and 6.5B active parameters with tool support; Tiny has 9.3B total and 2.7B active parameters for efficient dialogue. The ASR Tiny model has 8.6B total and 2B active parameters, with dialect and context support. inclusionAI reports support for seven TTS languages, a leading open-weight score of 1069 Elo in AA Controlled Voice, and a 4.41% average error rate for ASR Tiny across four recognition categories. Weights are available on Hugging Face and ModelScope, with launch-day support in SGLang and vLLM.

## Availability

Weights are available on Hugging Face and ModelScope, with launch-day support in SGLang and vLLM.
- Model released: Oct 19, 2026 · Date only
- Review updated: Oct 20, 2026 07:08 Tue · Beijing Time (UTC+8)

## Sources

- Source excerpt: https://x.com/AntLingAGI/status/2112210811997540475

Canonical entry: https://www.ailyre.com/source/?id=ant-group-inclusionai-ming-3-0&lang=en