No official torrent yet β requesting it tells the flock what to host next. You cannot seed a catalog-only page.
Cataloged from OpenMOSS-Team/MOSS-Audio-8B-Thinking
All models by OpenMOSS-Team β
MOSS-Audio is an open-source audio understanding model from MOSI.AI, the OpenMOSS team, and Shanghai Innovation Institute. It performs unified modeling over complex real-world audio, supporting speech understanding, environmental sound understanding, music understanding, audio captioning, time-aware QA, and complex reasoning. In this release, we provide four models: MOSS-Audio-4B-Instruct, MOSS-Audio-4B-Thinking, MOSS-Audio-8B-Instruct, and MOSS-Audio-8B-Thinking. The Instruct variants are optimized for direct instruction following, while the Thinking variants provide stronger chain-of-thought reasoning capabilities.
Understanding audio requires more than simply transcribing words β it demands the ability to perceive acoustic cues, recognize speakers and emotions, interpret environmental sounds, reason over temporal context, and handle complex multi-step inference. MOSS-Audio is built to unify these capabilities within a single model.
MOSS-Audio follows a modular design comprising three components: an audio encoder, a modality adapter, and a large language model. Raw audio is first encoded by MOSS-Audio-Encoder into continuous temporal representations at 12.5 Hz, which are then projected into the language model's embedding space through the adapter and finally consumed by the LLM for auto-regressive text generation.
Rather than relying on off-the-shelf audio frontends, we train a dedicated encoder from scratch to obtain more robust speech representations, tighter temporal alignment, and better extensibility across acoustic domains.
Using only the encoder's top-layer features tends to lose low-level prosody, transient events, and local time-frequency structure. To address this, we design a DeepStack-inspired cross-layer injection module between the encoder and the language model: in addition to the encoder's final-layer output, features from earlier and intermediate layers are selected, independently projected, and injected into the language model's early layers, preserving multi-granularity information from low-level acoustic details to high-level semantic abstractions.
This design is especially well-suited for audio understanding tasks, as it helps retain rhythm, timbre, transients, and background structure β information that a single high-level representation cannot fully capture.
Time is a critical dimension in audio understanding. To enhance explicit temporal awareness, we adopt a time-marker insertion strategy during pretraining: explicit time tokens are inserted between audio frame representations at fixed time intervals to indicate temporal positions. This design enables the model to learn "what happened when" within a unified text generation framework, naturally supporting timestamp ASR, event localization, time-based QA, and long-audio retrospection.
| Model | Audio Encoder | LLM Backbone | Total Size | Hugging Face |
|---|---|---|---|---|
| MOSS-Audio-4B-Instruct | MOSS-Audio-Encoder | Qwen3-4B | ~4.6B |  |
| MOSS-Audio-4B-Thinking | MOSS-Audio-Encoder | Qwen3-4B | ~4.6B | |
| MOSS-Audio-8B-Instruct | MOSS-Audio-Encoder | Qwen3-8B | ~8.6B | |
| MOSS-Audio-8B-Thinking | MOSS-Audio-Encoder | Qwen3-8B | ~8.6B | |
More model families, sizes, and variants will be released in the future. Stay tuned!
We evaluate MOSS-Audio on a comprehensive set of audio understanding benchmarks. Key results:
We recommend Python 3.12 with a clean Conda environment. The commands below are enough for local inference.
If your GPU supports FlashAttention 2, you can replace the last install command with:
Download the model first:
Then edit MODEL_PATH / AUDIO_PATH in infer.py as needed, and run:
The default prompt in infer.py is Describe this audio. You can directly edit that line if you want to try transcription, audio QA, or speech captioning.
Start the Gradio demo with:
If you want to serve MOSS-Audio with SGLang, see the full guide in moss_audio_usage_guide.md.
The shortest setup is:
If you use the default torch==2.9.1+cu128 runtime, installing nvidia-cudnn-cu12==9.16.0.29 is recommended before starting sglang serve.
[blocked]
Models in MOSS-Audio are licensed under the Apache License 2.0.
git clone https://github.com/OpenMOSS/MOSS-Audio.gitcd MOSS-Audio
conda create -n moss-audio python=3.12 -yconda activate moss-audio
conda install -c conda-forge "ffmpeg=7" -ypip install --extra-index-url https://download.pytorch.org/whl/cu128 -e ".[torch-runtime]"pip install --extra-index-url https://download.pytorch.org/whl/cu128 -e ".[torch-runtime,flash-attn]"huggingface-cli download OpenMOSS-Team/MOSS-Audio --local-dir ./weights/MOSS-Audiohuggingface-cli download OpenMOSS-Team/MOSS-Audio-Instruct --local-dir ./weights/MOSS-Audio-Instructpython infer.pypython app.pygit clone -b moss-audio https://github.com/OpenMOSS/sglang.gitcd sglangpip install -e "python[all]"pip install nvidia-cudnn-cu12==9.16.0.29cd ..sglang serve --model-path ./weights/MOSS-Audio --trust-remote-code@misc{mossaudio2026, title={MOSS-Audio Technical Report}, author={OpenMOSS Team}, year={2026}, howpublished={\url{https://github.com/OpenMOSS/MOSS-Audio}}, note={GitHub repository}}