Physix Frontier badgePhysix Frontier
PublishChinese | EN

Physix Frontier · News Briefing Card (IT Home · Oct 16, 2026)

Xiaomi Open-Sources MiDashengLM-Spatial Audio Model

KEY FACTS

  • Xiaomi released and open-sourced the MiDashengLM-Spatial model on October 15.
  • The model weights and inference code are published under the Apache 2.0 license.
  • The model is built on MiDashengLM-7B and adds spatial awareness and reasoning capabilities.
  • Xiaomi claims it is the industry's first open-source end-to-end spatial audio understanding large model.
  • Dasheng series technology has already been deployed in over 30 applications across smart homes and car cockpits.

KEY DATA

Over 30Number of deployed applications
7BBase model parameter count

PHYSIX OBSERVATION

Spatial hearing was a blank spot few had touched in audio large models, and Xiaomi's early open-sourcing means it holds the power to set standards. For developers, Apache 2.0 means commercial use is allowed, greatly lowering the barrier; for the industry, interaction competition across the full human-car-home ecosystem will shift from speech recognition to spatial perception, adding one more subject latecomers must catch up on.

Source: IT Home report