Dasheng

paper model
Xiaomi
Audio encoder and language model family. Original Dasheng encoder (1.2B params, 272K hours) evolved into DashengLM and MiDashengLM-7B for efficient audio understanding supporting speech, music, and acoustics.
Paper (arXiv, Dasheng)Paper (arXiv, DashengLM)GitHub

Outputs 3

Dasheng: Scaling Masked Audio Encoder Learning

paper

Original Dasheng audio encoder (1.2B params, 272K hours). Foundation for MiDashengLM and DashengTokenizer.

Paper (arXiv)

arXiv: 2406.06992

Dasheng-LM: Efficient Audio Understanding with General Audio Captions

paper

Research on efficient audio understanding using general audio captions.

Paper (arXiv)GitHub

arXiv: 2508.03983

MiDashengLM-7B

model

Efficient audio understanding model built on the "Dasheng" encoder, supporting speech, music, and acoustics.

GitHub
Architecture DENSE
Parameters 7B
audioarchitectureopen-weight