MOSS-Transcribe-Diarize

weights

open(0.9B safetensors checkpoint on the Hub, ungated)

data

closed(trained on in-the-wild recordings that are not listed or released)

code

partial(inference, serving and a minimal fine-tuning script)

license

Apache-2.0(code and the 0.9B weights)

api-only-tier

MOSS-Transcribe-Diarize Pro(a stronger model offered only through MOSI's online playground

The 0.9B model and its code are Apache-2.0 and download freely, but the training recordings are neither listed nor released and only a minimal fine-tuning script is published. A stronger Pro model is offered only in MOSI's hosted playground.

• https://arxiv.org/abs/2601.01554 recorded 2026-09-27

  Abstract: "Trained on extensive real wild data and equipped with a 128k context window for up to 90-minute inputs".

• https://huggingface.co/api/models/OpenMOSS-Team/MOSS-Transcribe-Diarize?expand[]=downloads&expand[]=cardData&expand[]=gated&expand[]=createdAt&expand[]=lastModified&expand[]=tags&expand[]=siblings recorded 2026-09-27

  Hub record: "gated":false and model-00000-of-00001.safetensors in the file list.

• https://huggingface.co/OpenMOSS-Team/MOSS-Transcribe-Diarize/raw/main/README.md recorded 2026-09-27

  Card: "MOSS-Transcribe-Diarize 0.9B is licensed under the Apache License 2.0."

• https://raw.githubusercontent.com/OpenMOSS/MOSS-Transcribe-Diarize/main/FINETUNING.md recorded 2026-09-27

  FINETUNING.md: "`finetune.py` provides a minimal fine-tuning workflow built on the Hugging Face Transformers `Trainer`."

• https://raw.githubusercontent.com/OpenMOSS/MOSS-Transcribe-Diarize/main/README.md recorded 2026-09-27

  README: "* 2026-07-09: Open-sourced MOSS-Transcribe-Diarize 0.9B."; "[MOSS-Transcribe-Diarize Pro](https://platform.mosi.cn/app/playground) is a stronger model with higher overall performance and is available through the online playground."

Hugging Face downloads of the 0.9B checkpoint within about three months of its open release.

• https://huggingface.co/api/models?author=OpenMOSS-Team&search=MOSS-Transcribe&limit=1000&expand[]=downloads&expand[]=createdAt&expand[]=tags&expand[]=gated recorded 2026-09-27

  Trailing-30-day downloads: OpenMOSS-Team/MOSS-Transcribe-Diarize 165,491.

Close to the best open recognizers on English accuracy while also diarizing long recordings, a step below Qwen3-ASR.

• https://huggingface.co/datasets/hf-audio/open-asr-leaderboard-results/resolve/d2c5b384deccdb82834f41aeaffcc618c00efa2f/english_short_latest.csv recorded 2026-09-27

  Results CSV row "OpenMOSS-Team/MOSS-Transcribe-Diarize,4.63625".

• https://raw.githubusercontent.com/OpenMOSS/MOSS-Transcribe-Diarize/main/README.md recorded 2026-09-27

  README: "an open-source SOTA end-to-end audio understanding model for long-form multi-speaker transcription, diarization, timestamps, and acoustic event awareness".