Sber presents "GigaAgent": a new era of autonomous AI assistants in Russia
Russian banking giant Sber has officially launched "GigaAgent" — a universal autonomous AI assistant designed to handle a wide range of everyday and work tasks. This move marks a significant escalation in the race for leadership in the AI agent segment, especially against the backdrop of recent similar announcements from Yandex. The market is clearly shifting 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 even the preparation of reports and presentations. The key difference from its predecessors is deep context memory and the ability to act without supervision at every step. The agent accumulates feedback, adapts to the user's individual work style, and most importantly, demonstrates initiative: it asks clarifying questions when necessary and can handle multiple tasks in parallel within a single semantic framework.
The developers highlight three pillars on which GigaAgent is built. The first is autonomy: the system operates around the clock, including in proactive mode, and honestly signals when data is insufficient. The second is self-improvement: the agent can analyze and rewrite its own code, recording each change as a commit in a built-in repository. The third is continuity: unlike most AI agents that operate within a single session, GigaAgent retains accumulated experience and interaction history, continuing work after a restart while taking into account already gathered context.
Technology and pricing
At its core lies the self-improving agent project Ouroboros, created by researchers at the AIRI Institute. On the Terminal Bench 2.1 benchmark with Anthropic's Opus 5 model, the agent demonstrates an impressive result of 86.97%.
The service is launched through the Agents Space platform from Cloud.ru. New users are provided with a starting grant of 4,000 rubles, allowing them to evaluate the platform's capabilities for free. Further usage is billed based on consumption of language models and infrastructure, with the final amount directly depending on the volume and frequency of use.
Estimates of development costs vary. According to experts, 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 around 100 people. For a company of this scale, this is a relatively small amount, especially considering that Sber previously released the base model, GigaChat 3, under a free license, which avoided the costs of training from scratch.
My analysis: The launch of GigaAgent is not just another product, but a strategic bet on shaping a new technological paradigm. However, as with any loud claims about autonomy, the key question remains measurable business metrics and control mechanisms. The market needs not promises, but evidence of the effectiveness and reliability of such agents in real-world use cases.