Yandex B2B Tech has announced a major update to Yandex AI Studio, its platform for building artificial intelligence applications. The update introduces new tools that allow businesses to create intelligent AI agents capable of performing complex tasks consistently.
The update also brings improvements to security, infrastructure, and pricing. According to Yandex, the new pricing model can reduce the cost of running AI agents by four times or more in certain scenarios.
Improved Infrastructure For AI Agents
The updated Yandex AI Studio is designed to support more advanced AI workflows. The platform now handles long contexts more effectively and supports multi-step processing for complex tasks.
Arthur Samigullin, head of Yandex AI Studio, said the company redesigned the platform’s infrastructure to improve reliability and scalability. The new system also manages repeated model calls, which helps maintain stable performance even under heavy workloads.
Samigullin said the combination of infrastructure upgrades and new pricing options should significantly lower the final cost of using AI agents for businesses.
Reasoning Agents Powered By DeepSeek-V3.2
Yandex AI Studio now enables developers to create AI agents using the DeepSeek-V3.2 model. The model offers stronger reasoning capabilities and can maintain longer context while generating consistent conclusions.
These AI agents can also interact with external applications while processing information. This allows them to perform tasks and respond more quickly.
Reasoning agents are designed for complex workflows. Businesses can use them to compare supplier tenders, investigate technical incidents, or manage software development tasks.
For example, they can run automated tests, create repositories, and deploy applications. The model has already been integrated into the neural network ensemble used by SourceCraft Code Assistant, which is part of Yandex’s SourceCraft development platform.
Yandex says running DeepSeek-V3.2 through its cloud platform is often simpler and more secure than deploying similar infrastructure independently. The platform processes generative AI data in Russia and complies with domestic and international security standards.
New Security Tools For Enterprises
The update also introduces additional security features designed for large organizations such as banks and industrial companies.
Yandex AI Studio can now be connected through a dedicated secure channel. This allows companies to use AI agents inside their corporate networks without requiring internet access.
Requests sent to the AI system are not logged, which means sensitive data remains inside the company’s infrastructure.
The platform is also now fully available for on-premises deployment. Organizations can install Yandex AI Studio directly on their own infrastructure or use a hybrid setup.
In a hybrid model, confidential data can be processed locally while resource-intensive AI tasks are handled in the cloud through a secure connection. According to Yandex, this approach is already being used by a major industrial company and a Russian bank.
New Pricing Model Introduces Token Types
The update also introduces a new pricing structure that includes two additional token types.
The first type is cached tokens, which apply when the same information is reused across requests. The second type is tool tokens, which are used when the model receives data from external services or applications.
These changes are intended to make pricing easier to track and help companies reduce costs for high-volume tasks such as searching files or the web.
Yandex says that for some neural networks, cached and tool tokens will cost four times less than standard input and output tokens. This pricing approach aims to make large-scale AI deployments more affordable for businesses using Yandex AI Studio.
