Xiaomi ai model: Chinese tech giant Xiaomi has introduced a new AI voice model. Midashenglm-7B, which is not just an experiment running in the lab, but now in the devices of real life. This model is already active in smart home systems and cars in China and the biggest thing has been launched as an open source.

Open door for developers

This model of Xiaomi has been issued under Apache 2.0 license i.e. any developer or company can use it in research and commercial projects, without any obstruction. Xiaomi’s intention is not only to show technical strength but to create a strong developer community that can give him an edge in the coming AI round.

It also understands the atmosphere but also AI

The biggest feature of midashenglm-7B is that it is not limited to just voice. This model can also recognize background music, surrounding environment sounds and other sound signals. The credit goes to Xiaomi Dasheng audio encoder and Alibaba’s Qwen2.5-Aomni-7B decoder’s mail which has equipped this AI with multiple abilities. It can also identify voices like AI Clap or Snap as a command.

Ready for use in both home and car

Xiaomi claims that this AI model is currently handling more than 30 smart features. In smart homes, it is being used in features like 24/7 sound monitoring and alert for unexpected voices. In cars, it controls the system through voice commands and if a user is learning a new language during the journey, it also gives him real-time pronunciation feedback. Some devices also include underwater wake-up mode which is active with just sound signals without touch.

Fast, light and smarter

Xiaomi says that the response speed of midashenglm-7B is very fast and it can handle 20 times more requests simultaneously than other models, that too without spending more memory. This means that now to run AI features, neither expensive server nor dependence on the Internet.

Also read:

Government’s strictness on Deepfake! New law coming to save digital identity