Sber presents "GigaAgent": a new round in the race of autonomous AI agents in Russia
The Russian artificial intelligence market is entering a new phase. Sber has officially launched "GigaAgent" — a universal autonomous assistant capable not only of performing routine tasks but also of self-learning. This move is a direct response to similar announcements from competitors, particularly Yandex, signaling the start of an intense battle for leadership in the AI agent segment.
From chatbots to digital partners
We are witnessing a fundamental shift: simple conversational models are giving way to systems that plan and execute multi-step processes. "GigaAgent" is a vivid example of this evolution. It manages calendars, works with documents, handles correspondence, searches for information, and prepares reports. Unlike traditional chatbots, the new agent remembers context and does not require control at every step, accumulating feedback and adapting to the user's individual work style.
Three key features of "GigaAgent"
The developers highlight three distinctive traits. First, autonomy: the agent operates around the clock, initiates clarifying questions, and can reasonably object when there is an error in reasoning. Second, self-development: it can improve its own code, recording every change in the repository — this is no longer just a tool, but a system with elements of reflection. Third, continuity: unlike most analogues, "GigaAgent" preserves experience and interaction history, continuing work after a restart while taking accumulated context into account.
Technological foundation and economics
At its core lies the Ouroboros project, created by researchers at the AIRI institute. On the Terminal Bench 2.1 benchmark, the model shows an impressive result — 86.97% for Opus 5 from Anthropic. The launch is available through the Agents Space platform from Cloud.ru, and new users are provided with a starting grant of 4,000 rubles to test the capabilities. Further usage is billed according to the cost of language models and infrastructure.
Expert estimates vary: according to analysts, investments in orchestration, connectors, interface, and quality assessment systems could have ranged from 1 to 3 billion rubles over a year and a half, with the project team numbering about 100 specialists. At the same time, Sber does not disclose exact figures, noting that the base model GigaChat 3 was already released under a free license, which significantly reduced costs.
My view: Claims about "self-development" sound ambitious, but what will be critically important is not marketing, but measurable business metrics and clear control mechanisms. The market has already seen many loud promises, so the success of "GigaAgent" will be determined not only by its technical characteristics, but also by its ability to truly integrate into work processes and deliver measurable value to businesses.