Xiaomi has announced the launch of a new open source audio model, named Midashenglm-7B, dedicated to enhancing user experiences in the fields of smart cars and home appliances.
A qualitative shift that exceeds the written texts
The new model represents a remarkable development in artificial intelligence tools that are no longer limited to textual treatment, but also expanded to include immediate and accurate audio interaction, which opens wider horizons for more flexible applications in daily life.
Technical integration with the Ali Baba model
Midashenglm-7B is based on the primary audio model of Choumi, and it comes with the QWEN2.5-omni-7B open source model developed by Ali Baba, in a step that enhances the treatment strength and the diversity of use cases.
Exceptional performance and high efficiency
According to the specialized Xiaomitime website, the model recorded amazingly over 22 general standards, outperforing its competitors in terms of response speed and treatment efficiency.
The first time time – the time to respond with the first word – is 25% less than the average time when similar artificial intelligence solutions.
An unprecedented ability to coincide
Midashenglm-7B has the ability to treat simultaneous operations by 20 times compared to traditional models, without the need for an additional capacity of memory, which makes it ideal for environments that require immediate and effective performance.
Training on open source data
Xiaomi has trained the new audio model on general data available, which enhances the transparency of performance and supports the developer community in developing easily integral audio intelligence tools with various smart systems.
