Demand for Chinese AI models surged in Russia. They’re cheaper and easier to pay for.

Clients of Russia’s major business-focused cloud platforms sharply increased their use of Chinese large language models in 2026, the Russian business daily Kommersant reported, citing market participants.

Representatives of the Yandex AI Studio platform told Kommersant that Yandex’s own models accounted for only 54% of tokens processed in the first half of the year. Alibaba’s Qwen took second place with 21.2%, followed by DeepSeek at 13.5%. GPT-OSS, an American model from OpenAI with open parameters, accounted for another 7.5%.

Artur Samigullin, the platform’s head, said overall use of Chinese models has grown more than fivefold compared with last year.

On MWS Cloud, a platform owned by MTS, use of Chinese AI models grew 11-fold in the first half of the year compared with all of 2025. Qwen led the platform, followed by GLM from the startup Z.ai, with Kimi from Moonshot AI rounding out the top three.

Cloud.ru, another platform that was part of Sberbank’s ecosystem until 2022, reported that overall token consumption grew 18-fold in less than eight months. Its representatives did not provide exact figures but said “a significant portion of demand is for Chinese models.”

Market participants said the growing use of Chinese AI models — not just in Russia — reflects how closely their capabilities have caught up with American models. Lower prices at comparable quality are also a factor. Chinese models are also easier to pay for in Russia, since sanctions have kept many Western companies from officially operating in the country.

The sharp rise in popularity of Chinese AI models worldwide began in early last year with the launch of the DeepSeek chatbot. Gradually, in terms of quality, they began to approach more expensive systems from Western developers. Currently, in terms of download numbers, Chinese developments lead the open-model segment.