Russia’s Sberbank rolls out GigaChat to take on ChatGPT

While Silicon Valley dominates the AI conversation, Russia’s biggest financial institution is jumping into the ring with its own generative model.

Russian lender Sberbank on Monday announced the launch of GigaChat, a new chatbot designed to rival OpenAI's ChatGPT. The chatbot is currently rolling out in an invite-only testing phase.

A native Russian alternative

Sberbank is pitching GigaChat as a worthy contender to ChatGPT, pointing to one primary edge: native Russian language fluency.

The bank claims this ability to communicate intelligently in Russian sets GigaChat apart from foreign neural networks, which often struggle with cultural nuance and grammar in non-English datasets.

Russia is not the only player taking this localized approach. In China, tech giants have rolled out their own contenders, including Baidu's Ernie and Alibaba's Tongyi Qianwen, both emphasizing native language capabilities. Even Elon Musk recently entered the fray, announcing plans for TruthGPT as a truth-seeking alternative.

Driven by sanctions and tech sovereignty

It might seem strange for a retail bank to develop a cutting-edge large language model, but Sberbank has been pouring money into tech for years.

According to Reuters, that investment has been driven by a push to reduce Russia's dependence on foreign imports.

That drive has turned urgent following sweeping Western sanctions and severe export controls. Sberbank’s AI ambitions are less about consumer gimmicks and more about securing critical technological sovereignty.

Why it matters

For Russian-speaking users and developers inside the country, GigaChat could offer a functional alternative in an environment where access to Western platforms is increasingly cut off or legally complicated. That said, it is only in closed testing right now, so it is not something everyday users need to scramble to try.

On an industry level, this launch reinforces the rise of sovereign AI. Large language models are quickly becoming geopolitical assets, forcing nations to develop local equivalents rather than rely on American tech stacks.

We think leaning into local language optimization is the smartest move Sberbank could make. But the elephant in the room is hardware: with strict export restrictions on cutting-edge chips, it remains to be seen whether Sberbank has the compute infrastructure required to keep training and scaling GigaChat over the long run.