Sber presents "GigaAgent": an autonomous AI agent that learns and develops on its own
The AI agent market in Russia is entering a new phase. Sber has officially launched "GigaAgent" — a universal autonomous assistant designed to handle both routine and complex work tasks. This move is a direct response to intensifying competition in a segment where Yandex has already announced similar solutions. We are witnessing a transition from simple conversational chatbots to full-fledged digital partners capable of planning and executing multi-step processes.
Assistant for routine tasks
GigaAgent takes over calendar management, document handling, correspondence, information search, and preparation of reports with presentations. The key difference from classic chats is its ability to remember context and act without supervision at every step. The system accumulates feedback and adapts to the user's individual work style, showing initiative: it asks clarifying questions and simultaneously solves several tasks within a single semantic field.
The developers highlight three key properties of the new product. First — autonomy: the agent operates around the clock, in a proactive mode, reasonably objects when data is insufficient, and honestly signals errors. Second — self-improvement: GigaAgent is capable of improving its own code, recording each change as a commit in a built-in repository. Third — continuity: unlike most agents that work within a single session, this assistant retains accumulated experience, decision history, and interaction context even after a restart.
Technology and price
At its core lies the self-improving agent project Ouroboros from the research team at the AIRI Institute. The results are impressive: on the Terminal Bench 2.1 benchmark for Anthropic's Opus 5 model, quality reaches 86.97%. The launch is carried out through the Agents Space from Cloud.ru, where new users receive a starting grant of 4,000 rubles for free platform testing.
The volume of investment in the project is not disclosed, however, according to estimates from independent experts, the costs for orchestration, connectors, interface, and quality assessment system could have ranged from 1 to 3 billion rubles over a year and a half. The project team numbers about 100 specialists. At the same time, the base model GigaChat 3 has already been released under a free license, which significantly reduced costs.
My view: GigaAgent is not just another chatbot, but a serious bid for leadership in the agentic AI race. However, behind the loud promises of autonomy and self-improvement, there must be measurable business metrics and clear control mechanisms. The market will evaluate not words about capabilities, but the real economic efficiency and reliability of such systems in production scenarios.