Yandex challenges GPT: new neural network for business solves problems five times cheaper than Western counterparts

Yandex has launched a new language model, Alice AI LLM Flash, optimized for business tasks. At Yandex B2B Tech, the neural network provides high response speed and reduces data processing costs by almost five times compared to previous solutions. The model is focused on working with texts and documents – a segment that covers about 60% of all b2b requests to Yandex's neural networks.

Developers state that in 56% of cases, Alice AI LLM Flash outperforms GPT-5.4 mini in the quality of solving business problems. In dialogue scenarios, the Russian model is better in 73% of cases, in text summarization and structuring tasks – in 66%, and in data retrieval from files and knowledge bases – in 61%. At the same time, it retains all the key advantages of the flagship line: security of processed data and stable operation in Russia.

The launch of such a model is not just import substitution, but the creation of a tool capable of competing in quality and price with leading global counterparts. For Russian businesses, this means the ability to automate call centers, content moderation, and document management without dependence on foreign platforms and with a predictable budget.