Translated with AI from the Russian original.

The second generation of the large language model Cotype Pro is focused on creating corporate AI assistants. According to developers from MTS, the new version is 40% faster than the previous one and 50% more accurate in processing long texts - up to 128 thousand tokens (about 900 thousand characters).

At the same time, the classification accuracy remained at a high level - 87%. Cotype Pro 2 also handles reasoning better, which is useful for financial analysis and scientific research. In addition to basic tasks, Cotype Pro 2 supports "role-playing" - conducting a dialogue on behalf of a specialist, which is useful for customer support and employee training.

Cotype Pro 2 is adapted for more than 100 corporate scenarios and can be deployed on the internal infrastructure of companies. In 2025-2026, based on it, the MTS AI team plans to release a series of AI agents for the public sector, banks, retail and other areas.

In addition to basic tasks, Cotype Pro 2 supports "role-playing" - conducting a dialogue on behalf of a specialist, which is useful for customer support and employee training. The company also developed its own method for evaluating work with long texts, where the model showed a two-fold increase in quality compared to the previous version.

Now on the front page

Articles

Why even a PC case should be branded

In Russian conditions, the popularity of PCs assembled by small local companies or in-house corporate IT departments continues to grow rapidly. Such solutions have historically been more popular in Russia than in other developed markets, but today new nuances have emerged in the segment's evolution, which we discussed with Karim LASKRI, Business Director of FSP Group in the CIS region.

News

Storage you can trust in the background: Basis SDS 2.3 released with scheduled diagnostics and an observer role

A storage system is the only infrastructure layer where an error cannot be “rolled back”: neither a reboot nor a configuration rollback will restore lost data. That is why users expect predictability from software-defined storage rather than just new features—ensuring background processes finish what they started, and administrators can see the cluster status not only “right now,” but also over time.