Russia's largest IT holding is testing the sensational Chinese AI model DeepSeek

T1 IT Holding has begun testing DeepSeek technology for internal needs. TAdviser was told about this at the end of January 2025 by the general director of the T1 IT holding Alexey Fetisov. According to him, the technology is tested, for example, for code generation tasks, tests, for use in digital assistants, for logging meetings held on VKS, text analysis, etc.

The Chinese neural network DeepSeek made a lot of noise and at the end of January even came out on top of the top of the most downloaded applications for the iPhone began in the United States, and then in Russia. Among the reasons for the increased attention is that two DeepSeek models - the V3 neural network and the "reasoning" R1 - in tests bypassed competitors from OpenAI, Meta (an organization whose activities are recognized as extremist and banned in the Russian Federation) and Google.

Despite the fact that the basic languages ​ ​ for DeepSeek are English and Chinese, it works surprisingly well with texts in Russian, including the analysis and generation of jokes and phraseological turns, notes Alexey Fetisov.

Models such as DeepSeek have a great future, he is sure. For the model to become smarter, it needs to be further trained, and this requires a lot of appropriate equipment, which means additional costs for the business. DeepSeek is less hardware demanding, meaning models can be trained for less money. Actually, that's why NVidia shares fell. In addition, some pre-existing tests show that DeepSeek allows you to train models faster than known Western counterparts.

In general, the use of AI to generate code is not a new task, and for it the IT-Business, including T1, have been use various AI technologies for quite some time. In "T1" now a significant part of the code is already generated using AI. The quality, speed and efficiency of development thanks to AI is increased, which provides competitive advantages for the company in the market, TAdviser explained in T1.

𝓲

Фото: Freepik

Code generation products and services today are one of the most important trends in the use of generative AI. In part, they help address the shortage of skilled programmers in the labor market, reducing the time they spend on simple routine tasks. Also, AI tools can significantly reduce the time required to write code, which allows developers to quickly bring projects to completion.

But AI is not yet able to generate complex, high-level code that requires a deep understanding of the subject area and architectural solutions. As a result, AI will not replace experienced programmers in solving complex problems that require deep knowledge and creative approach, says the CEO of T1.

According to Alexei Fetisov, in addition to using AI for internal needs, the holding is actively introducing AI into its products, trying to increase the efficiency and quality of the services provided. For example, AI algorithms analyze user behavior and help offer personalized recommendations in Note Modus products, which increases customer satisfaction and increases sales of T1 partners. And in the Note Dome ecosystem, AI helps automatically detect the threats of cyber attacks, which allows you to respond to incidents faster.

We see great prospects for using AI in the future, from improving user experience to significantly improving the efficiency of business processes. However, to achieve these goals, it is necessary to continue to invest in research and development, as well as to ensure a high level of security and transparency in the use of AI systems, "said Alexey Fetisov in a conversation with TAdviser.

Note that the anchor customer of T1, VTB, is also aimed at expanding the practice of using AI. So, in 2024, Deputy President - Chairman of the Board of VTB Bank Vadim Kulik predicted that by 2026 each VTB client will have a digital assistant. VTB also stated that as part of its new development strategy until 2026, artificial intelligence will appear in almost every banking process.