Neural networks will be able to process information within the device without sending data to the cloud or relying on an internet connection. A resident of "Technopolis Moscow" presented LinQ HX — the first Russian neural network accelerator for embedded systems. The development was showcased at the international industrial exhibition "Innoprom-2026".

The accelerator was created by the company "HighTech". It is based on a domestic microprocessor, and the module itself is assembled from Russian components and does not require foreign licenses.

According to the developers, LinQ HX performs up to 30 trillion operations per second. It can simultaneously support the operation of several neural networks, and the processing delay is between 1.5 and 2.3 milliseconds. The device consumes no more than 40W.

In the future, such an accelerator is planned for use in video surveillance systems, robotics, and AI-based medical solutions. This includes diagnostics and patient monitoring. However, the technology is not yet applied in urban infrastructure.