Sber presents "GigaAgent": a new turn in the race of autonomous AI agents in Russia
Russia's largest bank is officially launching "GigaAgent" to the market — a universal AI assistant capable of working autonomously and self-learning. This move marks an escalation of competition in the intelligent agents segment, where Yandex has already announced a similar solution.
From chatbots to digital partners
We are witnessing a fundamental paradigm shift: the market is moving away from simple conversational interfaces toward full-fledged digital partners that not only generate text but also plan and execute multi-step business processes. "GigaAgent" is a vivid confirmation of this trend. Unlike traditional chatbots, the new agent can manage calendars, process documents, handle correspondence, and prepare reports without requiring oversight at every stage.
The key difference lies in an architecture built on three principles. First, full autonomy: the agent operates around the clock, shows initiative, asks clarifying questions, and can reasonably challenge incorrect input. Second, self-development: the system can analyze and modify its own code, recording each change in the repository as a checkpoint. Third, continuity: unlike most AI solutions that operate within a single session, "GigaAgent" retains context, interaction history, and accumulated experience even after a restart.
Technological foundation and project economics
At its core is the open-source project Ouroboros from the AIRI research institute. The results are impressive: on the Terminal Bench 2.1 benchmark, performance quality with Anthropic's Opus 5 model reaches 86.97%. This is a serious claim to leadership among domestic developments.
The launch is carried out through the Agents Space platform from Cloud.ru. New users receive a starting grant of 4,000 rubles to test the capabilities, followed by pay-as-you-go pricing for language models and infrastructure usage. According to independent experts' estimates, the total investment in creating the agent itself — orchestration, connectors, interface, and quality assessment system — amounted to between 1 and 3 billion rubles over a year and a half, with the project team comprising about 100 specialists.
For a bank of this scale, this is a relatively modest investment, especially considering that the base language model (GigaChat 3) did not have to be trained from scratch — it was previously released under a free license. However, it is important to understand: every task performed by the agent has a real cost, since it accesses multiple systems, makes mistakes, and asks for clarification — all of which requires computational resources.
My analysis: Given that Sber has already announced a lineup of crypto products and AI banking for businesses, the launch of "GigaAgent" is not an isolated product but part of a systematic strategy to integrate autonomous AI into all financial and corporate services. Investors and developers should closely monitor how these agents will be monetized and which business metrics will actually demonstrate their effectiveness.