Sber presents "GigaAgent": an autonomous AI agent that learns and develops on its own
The AI agent market is entering a new phase, and Russian technology giants have no intention of staying on the sidelines. Amid intensifying competition, Sber has officially launched "GigaAgent" — a universal autonomous assistant designed to handle both everyday and complex work tasks. This is a direct response to similar developments emerging from foreign and domestic competitors, underscoring the strategic importance of this direction.
From chatbot to digital partner
The key difference between "GigaAgent" and conventional conversational models is its ability to act without control at every step. This is no longer just a tool for generating text, but a full-fledged digital partner that can plan and execute multi-step processes. The agent can manage a calendar, work with documents, handle correspondence, search for information, and prepare reports with presentations. Moreover, it does not simply execute commands but shows initiative: it asks clarifying questions and raises reasoned objections if it spots an error in logic.
Three key properties: autonomy, self-improvement, and continuity
The developers highlight three fundamental differences between "GigaAgent" and existing solutions. First, full autonomy — the system operates around the clock in the background and honestly reports when data is insufficient. Second, self-improvement: the agent can analyze and rewrite its own code, recording each change as a checkpoint in a built-in repository. Third, continuity — unlike most AI systems that work within a single session, "GigaAgent" remembers past tasks and decisions, preserving context even after a restart.
Technology foundation and cost
The project is based on a development by the AIRI research institute — the self-improving agent Ouroboros. Its effectiveness has been confirmed on international benchmarks: on Terminal Bench 2.1 for Anthropic's Opus 5 model, quality reaches an impressive 86.97%. For now, the service is available through the Agents Space platform from Cloud.ru, where new users receive a starting grant of 4,000 rubles to test the capabilities.
The financial side of the matter is of interest. Although Sber does not disclose the exact investment amounts, independent experts estimate the cost of creating the agent — orchestration, connectors, interface, and quality assessment system — at 1–3 billion rubles over a year and a half of work by a team of about 100 specialists. This underscores the seriousness of the bank's intentions, which has already announced plans to develop AI banking for businesses.
My analysis: The launch of "GigaAgent" is not just technological PR, but a signal of market maturity. However, behind the loud claims of self-improvement and autonomy, there must be measurable business metrics and clear control mechanisms. Without this, the AI agent race risks turning into a competition of flashy press releases rather than real products. That said, Sber's bet on AI is obvious and systematic, which in the long term could radically change the landscape of digital services.