Deeray cut costs 30 times by using AI services Cloud.ru

2025: Moving Product Environment to Cloud.ru

IT company Deeray - developer of the AI platform for speech and text analytics for business - has moved the product environment to Cloud.ru. Thanks to the migration, the company has reduced the cost of language models by 30 times and has a reliable infrastructure for processing 4 million minutes of speech every month. The company Cloud.ru announced this on November 21, 2025.

Deeray's solution does not just translate speech into text, but reveals the context of the entire dialogue. For example, it can determine whether the issue of the person who turned to the organization's call centers has been resolved. Deeray uses ensembles of different LLMs instead of one universal model, as competitors do. The configuration is selected taking into account the economic feasibility and tasks of the project - from semantic analysis to multimodal data processing. Such an architecture requires the cost-effectiveness of GPU resources and the stability of the infrastructure.

Deeray has reduced the cost of using popular open source models using the Cloud.ru Evolution Foundation Models service. For comparison, one of the models presented in it (Qwen 2.5) costs the company 20 rubles for 1 million output tokens, while competing solutions cost about 600 rubles. The average Deeray prompt contains 6-8 thousand tokens in structured form, excluding system prompts, which gives about 30 times savings on each request.

Also, access to powerful GPU servers Cloud.ru allowed Deeray to cope with a sharp increase in customers from medium and small businesses. As of November 2025, the platform consistently handles 4 million minutes of speech and 1.5 million monthly chats, while maintaining the opportunity for growth. In addition, due to further training of models, the company managed to significantly reduce the number of speech recognition errors and in a number of domain areas to achieve an indicator of 1.5% against 25-29% among competitors.

For businesses whose products are built on AI, the price of open language models and graphics accelerators was the determining factor. In the case of Cloud.ru, the cost of resources turned out to be the most affordable on the Russian market, - said the founder and CEO of Deeray Dmitry Shcherbakov.

We try to improve processes and improve the economy of creating products without making them more expensive. We make AI products available to encourage companies to move from working with technology in R&D format to commercial consumption, "said Alexey Ri, leading Cloud.ru sales manager.

Since 2018, Deeray has been among the representatives of the Russian market using their own large language models instead of ready-made solutions. Since its launch, the company has implemented more than 400 projects.