Xiaomi open-sources MiDashengLM-Spatial, an audio model that hears both what and where

Xiaomi has open-sourced MiDashengLM-Spatial, a new member of its Dasheng audio model family designed to understand not only what a sound is but also where it comes from. As Pandaily reported on October 16, the company published the full model weights and inference code under the Apache 2.0 license, which permits both research and commercial use.

The model is built on MiDashengLM-7B, the sound-understanding large model Xiaomi open-sourced earlier, and adds a spatial audio encoder called Spatial-Dasheng. According to the technical report, a hierarchical semantic-to-spatial conditioning module captures semantic and spatial information separately and then merges them, producing a single end-to-end model that can describe speech, environmental sounds and music while reasoning about a source's direction, distance and movement.

The Hugging Face listing puts the model at roughly 8 billion parameters. Spatial cues depend on multi-channel input: the model card says it was trained on binaural, two-channel audio, and that mono input still runs but carries no directional information. Xiaomi says general audio abilities are preserved, with the research team reporting performance on basic audio tasks comparable to large audio-language models of around 8 billion parameters.

Xiaomi describes MiDashengLM-Spatial as the industry's first open-source end-to-end spatial audio understanding model. The paper makes a similar claim qualified with "to our knowledge," and the claim comes from Xiaomi and its researchers rather than independent assessment.

Over the past two years Xiaomi has open-sourced the Xiaomi Dasheng sound foundation model and MiDashengLM-7B. The company says technology from the Dasheng series already powers more than 30 applications across its smart home and in-car cockpit products, covering speech, ambient sound and music perception, a figure that refers to the existing series rather than the new model.

For MiDashengLM-Spatial itself, Xiaomi says it will be rolled out gradually in smart home and car cockpit scenarios, where knowing where a voice or an alarm is coming from could let assistants respond more precisely. No timeline or specific products have been named.

The release includes a project demo page, the arXiv technical report, evaluation code on GitHub and model weights on Hugging Face. Spatial audio understanding has long been constrained by data and evaluation standards, so whether a commercially licensed, end-to-end open model can be reproduced and benchmarked quickly by the community is the thing to watch.