InclusionAI releases Ming-3.0-audio-flash, a multilingual audio-language model
inclusionAI has open‑sourced ming-3.0-audio-flash, a native audio‑language model that processes spoken input together with text instructions. The model handles Chinese and English speech, including dialects, accents, noisy environments and multi‑speaker recordings, and can perform long‑form audio understanding, acoustic scene analysis, and reasoning over audio content. It also offers optional…
Key points
- inclusionAI released ming-3.0-audio-flash, supporting speech recognition, audio reasoning, and optional speech synthesis for Chinese and English.
- Model can be deployed with vllm‑omni on eight 80 GB H800 GPUs for 128K context or with sglang on four GPUs for text output.
- The GitHub repository logged 22 downloads last month and is distributed under the MIT License.
The repository provides a Python example that runs the model with the Transformers library, as well as deployment scripts for vLLM‑Omni (using eight 80 GB H800 GPUs for a 128K serving profile) and SGLang (four GPUs for text‑only output). The project is released under the MIT License, and the GitHub page recorded 22 downloads in the last month. inclusionAI notes that users should review the code before enabling remote execution flags, reflecting a focus on security and trust.
The release adds a new multimodal capability to the Hugging Face ecosystem, allowing developers to build applications that understand and respond to audio in multiple languages without needing separate speech‑to‑text and language models.
inclusionAI/Ming-3.0-audio-flash · Hugging Face
huggingface.co · 19 October 2026
Loading the full article…
This text was published by huggingface.co. It is reproduced here with attribution so you can read it in full; the rights remain with the publisher. Read it at the source ↗
Coverage and discussion
2sources- Reddit discussionreddit.com
- Reddit discussionreddit.com
- inclusionAI/Ming-3.0-audio-tiny · Hugging FacePrimary source · huggingface.co ·
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Generative AI & Models
All →- MiniMax releases M3.1 Flash model at $0.15 per 1M input tokens · 2 src
- Anthropic switches from git to internal Scion as AI writes over 30,000 code changes daily · 1 src
- doubleAI releases debut code execution model and raises $220 million · 1 src
- DeepSeek releases V4.1-Flash-1017 update with lower cost and higher DeepSWE score · 25 src
- Claude Fable 5 leads five‑way AI trading showdown, beating Mainline AI and GPT-5.6 Sol · 1 src
Comments
via GitHub Discussions