Yandex Cloud Product Digest (April 2026)

Infrastructure and network

A big update to Yandex Cloud Video

The service has new features for protecting and processing content quickly:

• overlaying logos in videos for branding and protection from copying
• companies will be able to manage broadcast delay and track signal stability in monitoring according to schedules in real time
• speeding up playback by showing an original video during parallel transcoding
• customer support will help customers quickly transfer video archives from other hosting sites
• the player can now be easily integrated into LMS educational systems using the Moodle plugin — this is especially important for universities
• monetization for online broadcasts with YAN.

More details about the update were described in a separate news article.

S3 Inventory launched

The new functionality for auditing data in Yandex Object Storage backups helps you get a tabular upload of metadata for objects in the bucket, without manual binding via the S3 API and List and Head operations. The report provides a list of objects with key information: size, creation date, storage class, MD5 hash, versioning, access settings, and other parameters.

Working with S3 Inventory generates a report during periods of reduced load, speeding up data analysis and reducing the number of operations to storage — this leads to savings in resources and costs.

New configuration options added to Yandex Cloud CDN

You can now set a security profile for a resource by selecting a preset that contains a specific set of TLS enabling ciphers.
In addition, it is now possible to set a custom SNI when contacting the user’s source. The functionality is useful:

• when using shared hosting, when one IP address serves several domains
• to correctly process SSL/TLS certificates when working with multiple-domain configurations.

For location rules, we added the ability to reuse another source without inheriting from the main configuration. This is relevant for scenarios where you need to redirect a specific URL to another source.

Notifications in Yandex Messenger

We have expanded the notification capabilities in the Yandex Cloud console. Now, in addition to current channels, you can receive notifications via Yandex Messenger. Technical notifications, monitoring, and billing alerts will be available. We talked about how to set them up.

For escalations and alerts from Monium, create a channel in the Notification methods and add recipients to the alerts sent to Telegram. In Escalation policies, add a step with the Cloud Yandex Messenger method and specify recipients — the changes will be automatically applied to all related alerts.

Yandex Cloud Backup

We significantly accelerated the work of the backup cluster and increased backup creation speed. In some cases, this can reducde backup creation times threefold.

Data platform

Yandex Managed Service for PostgreSQL

We’ve launched a new performance diagnostic screen — now finding problematic queries (and more) is easier. We talked about how to use it at our webinar. We also shared our experience on how to troubleshoot Postgres performance issues at a performance meetup.

In addition, it is now possible to customize your backup schedule by specifying the frequency of storage duration and withdrawals.

Yandex Managed Service for MySQL^®

• In the past, customers who set Master priority might find themselves in a situation where, after upgrades or maintenance operations, the Master ended up on a lower priority host, e.g., in a non-priority availability zone. The system now automatically switches Master back to the highest priority host.
• We added protection against deleting a database similar to a cluster. Now the risk of accidentally losing data is reduced.
• We’ve added the ability for users with the mdb_admin role to see which users they’ve created. Just check the tables available — there you will see a view with a list of users of the mysql.v_user, mysql.v_db and mysql.v_procs_priv types.

Yandex Managed Service for Valkey^™

Stable FQDNs for clusters have appeared: three c-{cid}-valkey{0,1,2}.db.yandex.net type CNAME records have been created that lead to cluster hosts distributed across different availability zones and shards. Records are automatically recalculated when topology changes (adding or removing hosts or shards) and switched when the host is unreachable — this provides a stable entry point regardless of changes in the cluster.

Yandex Managed Service for OpenSearch

It is now possible to seamlessly change a disk from HDD to network-ssd. This allows you to be more flexible when choosing how to change disk types.

Yandex MPP Analytics Engine for PostgreSQL

We held a webinar to explain how to use our new tool for more convenient and easier use of Greenplum^® / Cloudberry^™, the command center.

Data Catalog

• We supported Yandex DataLens as a source in Data Catalog — you can now build connections (lineage) from database to a dashboard.
• Version comparison. The catalog now allows you to compare different versions of schemas and descriptions.
• Resource status in lineage. You can see the current status of a resource directly in the lineage, e.g., whether a transfer from DataTransfer is working or an error has occurred. This makes it possible to catch errors and problems in data delivery more quickly.

Yandex DataLens

A new format for dashboard settings and reports

The settings have moved from the pop-up window to the sliding panel on the right. You can now adjust the spacing between charts. Visual changes are visible immediately after application, making it easier for you to customize dashboards and reports.

Activities on Editor charts

Now you can configure interaction with external systems directly from charts and selectors.

This can be used to:

• Send a request via API Connector to your service by clicking on a button, graph, or table row.
• Process a response and show the result in a pop-up window, toast, or update dashboard settings.

The function is available for Editor charts: tables, Gravity UI Charts, and selectors.

Artificial intelligence

Qwen 3.6-35B-A3B is now available in Yandex AI Studio

Qwen 3.6 will replace the previous version of the Qwen 3.5 multimodal language model.

The new version is significantly superior to its direct predecessor, especially in terms of agent coding and reasoning:

• Improvements in reasoning. Qwen 3.6 shows a significant increase in some benchmarks related to reasoning, especially in mathematics, logic, and multi-step problem solving.
• More accurate following of instructions. The model is better able to adhere to structured output formats, comply with restrictions, and more accurately process nuanced instructions without bias.
• Improving the quality of generated code. The increase in quality is especially noticeable in tasks with a large code base and when working with less common programming languages.

How to update a model

If you use the model via Yandex AI Studio UI, you don’t need to do anything else — the model will automatically update to the new version. If you’re using Qwen 3.5 through the Responses API, you’ll need to update model_uri.

Previously:

gpt://<folder_ID>/qwen3.5-35b-a3b-fp8

Now:

gpt://<folder_ID>/qwen3.6-35b-a3b

Until May 28, when referring to the previous model_uri, the new Qwen 3.6 model will be called by default. After May 28, support for the past model_uri will end. Qwen 3.6 costs the same as Qwen 3.5. More details on the pricing page.

Code Interpreter released in GA

The tool is now available to all users in the Yandex AI Studio interface.

You can:

• Create an AI agent and connect Code Interpreter to it.
• Upload new files for processing — in the tool settings or directly into a text request.
• Add files already downloaded from the vault.

Code Interpreter extends the capabilities of AI agents and provides the model with a full-fledged Python^® runtime built into the reasoning process.

Its key feature is iterative, when code execution is not a one-time action.

The model can consistently refine logic, correct errors, and complicate the solution, based on the results of previous runs.

Yandex AI Studio Academy launched

This is an educational project for everyone who wants to use artificial intelligence for their work and business.

It is suitable both for developers of AI solutions and for those who plan to incorporate neural networks into their workflows, e.g., preparing reports, market research or analytics. The materials will help you create AI agents even if you have no coding skills.

Several formats are available:

• Theoretical courses are practical materials about working with AI from leading Yandex AI Studio specialists and partners.
• Cookbooks are step-by-step practical instructions for creating AI solutions.
• Lab work is an interactive simulator where you complete a task and receive instant feedback from an AI assistant.
• The System Prompt Library offers instruction templates that will help your agents work more efficiently.

All SpeechKit features now available via the Yandex AI Studio SDK library

Yandex AI Studio SDK has added support for SpeechKit speech recognition and synthesis services: convenient, typed interfaces for working with the API and detailed examples are available.

In addition, the SDK implemented bidirectional streaming that emulates WebSocket interaction over the SpeechKit API. This makes it easier to develop voice interfaces and other scenarios where low latency and real-time data exchange are important.

Speech-to-Text (STT)

All STT API v3 features including:

• streaming speech recognition
• deferred recognition, including audio processing via S3 link
• classifiers
• analysis of individual speakers' speech and dialogs in their entirety
• LLM post-processing of recognition results.

Text-to-Speech (TTS)

Except audio templates, all the features of the TTS API v3, including streaming speech synthesis and use of Yandex SpeechKit Brand Voice.

SDK integration includes:

• a single dependency on the yandexcloud package without version conflicts
• integration with other parts of Yandex AI Studio SDK
• the option to directly transfer STT results to the OpenAI-compatible API.

Create search indexes natively using Yandex AI Studio SDK tools

The Yandex AI Studio SDK now has the Vector Store CLI console utility, which allows you to create Vector Store search indexes based on files from various sources, including:

• Local file system
• Atlassian Confluence;
• MediaWiki engine systems
• Yandex Object Storage and other S3-compatible storages.

How to install the Vector Store CLI in the correct configuration:

• Basic installation with support for local files and Atlassian Confluence: pip install yandex-ai-studio-sdk
• With Amazon S3 support and Object Storage: pip install "yandex-ai-studio-sdk[cli-s3]"
• With MediaWiki support: pip install "yandex-ai-studio-sdk[cli-wiki]"
• All additional features: pip install "yandex-ai-studio-sdk[cli-wiki,cli-s3]"

Updated format of interaction with Realtime API

The old format will be supported until May 12. We recommend you update the integration in advance to maintain compatibility.

What’s changed:

• The API configuration is now more structured: the main parameters have been moved to a single session block.
• Audio, VAD, and modality settings are now set in an explicit and consistent structure.
• We simplified a number of events and calls, including creating a response.
• The format of function parameters and messages has changed: the structure is stricter.
• We updated the connection endpoint.

Updated Realtime API, with more convenient implementation of voice scenarios

Audio configuration

Now in the settings you can manage:

• recognition language: audio.input.languages
• voice role in synthesis: audio.output.role

This allows you to more accurately customize the voice agent’s behavior to the desired scenario, e.g., choosing a language for a particular market or setting the appropriate synthesis style for the selected voice.

For example:

"audio": {
  "input": {
    "format": {
      "type": "audio/pcm",
      "rate": IN_RATE
    },
    "languages": ["kk-KZ"],
    "turn_detection": {
      "type": "server_vad",
      "threshold": 0.5,
      "silence_duration_ms": 400
    }
  },
  "output": {
    "format": {
      "type": "audio/pcm",
      "rate": OUT_RATE
    },
    "voice": "zhanar",
    "role": "friendly"
    }
  }

In the example, languages is the recognition language, and role is the role for the selected speech synthesis voice.

Consumption details

We added the token consumption rate to usage. Now you can see:

• incoming tokens
• cached tokens
• outgoing text tokens
• outgoing audio tokens
• total consumption

For example:

"usage": {
  "input_tokens": 1019,
  "output_tokens": 7,
  "total_tokens": 1026,
  "input_token_details": {
    "text_tokens": 27,
    "cached_tokens": 992
  },
  "output_token_details": {
    "text_tokens": 5,
    "audio_tokens": 2
  }
}

Such detail helps better understand what makes up consumption and more accurately estimate the cost of voice scenarios.

Code Interpreter webinar

Sergey Narbut, Head of the Yandex AI Studio Development Team, and Evgeny Baranovsky, GenAI Project Manager, spoke at a webinar about the capabilities of modern AI agents and agency tools, what Code Interpreter is, how it differs from RAG and uploading files directly to LLM, and how it works from a technical point of view.

Launching the Community Heroes program in Yandex AI Studio

We want to recognize and support the contribution of everyone who helps develop Yandex AI Studio — sharing ideas, helping other users, and telling about the capabilities of our platform.

How the program works:

• We mark a Community Hero in the community with a special icon. We hope that the official title of Yandex AI Studio Community Hero will be useful in professional activities and will increase your external expert status.
• A Community Hero’s answers to community questions will be converted into Yandex AI Studio grants. Our expert will be able to give them to anyone or keep them for themselves (if they want to).
• To help our expert community grow, we will publish a Community Heroes Digest every month, where we rank community members by the number of useful answers in the community.

Wordstat API will move from preview to GA on May 18

The public version of the service will be available to all users. It can fully be used to solve your problems. The service will be provided in accordance with the following rates for calls using various methods:

• GetTop — 20.00 RUB/1000 requests
• GetDynamics — 20.00 RUB/1000 requests
• GetRegionsDistribution — 50.00 RUB/1000 requests
• GetRegionsTree — free

Read more about connecting the service in the documentation.

Security

Yandex SIEM launched

This is a cloud-based SIEM system for businesses that processes up to 3 million events per second and scales automatically, without manual capacity expansion.

The reliability and resiliency of the system has been proven by its operation on the scale of Yandex, as well as part of the Yandex Cloud Detection and Response service, which is already used by major customers.

One feature of the solution is the built-in multi-agent system for analysts that automates incident analysis and speeds up the response to threats. The pay-as-you-go subscription model transfers costs from capital to operating costs, reducing the entry barrier and allowing you to manage your budget flexibly. Subscription access removes the main barriers to implementation: high start-up costs, scaling difficulties, and long-term deployment.

Cloud Security Day 2026 held

The conference talked about new products and updates.

Together with SolidLab, they presented SolidWall AI Security Gateway, an AI-based application security solution. It will help companies protect digital products from cyber threats and leaks of personal and corporate data. The solution was developed on the basis of Yandex Cloud technologies, taking into account the expertise of SolidLab. It is available for deployment both within the client and using the cloud model. To get access, you must submit a request on the website.

Yandex Smart Web Security now protects against DDoS attacks at the L3 and L4 levels using Yandex technology. In addition, we’ve integrated a new Solid WAF module that combines negative and positive security models, analyzes user behavior to identify bots, and inventories the API.

And the Security Deck service now has a Vulnerability Management vulnerability management module — it ensures application security at all stages, from build to actual operation.
The module centrally scans applications in repositories at the development stage, as well as those that are ready when they are launched and running. Detected threats are automatically sorted by criticality, which makes it possible to quickly eliminate the most dangerous ones and effectively manage infrastructure risks.

Other news

• The Yandex Cloud Registry service is now available in Kazakhstan. It can store Docker^® images, Java^™ images, and Node.JS artifacts. Thanks to the registry, it is possible to significantly simplify the management of dependencies between project components.
• We published a case study on how Anjoy Research stops attacks using Yandex Smart Web Security and reduces risks.
• Sergey Nesteruk, responsible for the security of using artificial intelligence at Yandex Cloud, explained in an article on Habr how to turn an AI agent into a trusted member of the information security team.
• We participated in several events: at CISO FORUM 2026 with a stand and in discussions about information security performance metrics and AI agents in information security; at Cyberdom, in a discussion on cyber risk insurance; as well as in the organization and program of the 18th meeting of the Information Security in Industry club.

DevOps tools

SourceCraft spring release

Here are the main platform updates:

Cloud IDE

• SourceCraft Spaces is available for you to test if you requested a preview.

SourceCraft Code Assistant

• Custom AI skills are now available in the VS Code plugin.
• New MCP tools for Yandex Cloud have appeared in the VS Code plugin marketplace: for generative search, deployment, and other scenarios.

Security

• License scanning automatically identifies transitive dependencies in the project, finds related vulnerabilities, and checks licenses.
• The dependency map is an interactive graph of project dependencies with information about vulnerabilities, licensing risks and statuses that helps analyze them faster and plan updates.
• A separate page in the scan history shows a list of past scans.

Automations

• Cube parameters can now be reused from another repository’s YAML file.
• The values of the input parameters passed are now visible in a running process.
• The automation interface was updated — now the launch parameters can be viewed in separate cards or in a general list.

New tools and platform improvements

• Mapping commits to a profile helps to uniquely identify the author in the repository’s change history and link commits to a profile on the platform.
• Change suggestions support notifications, labels, and easy search.
• The new configuration visualizer draws parameters from YAML and validates them according to the scheme.
• The repository’s raw files are now available via direct links.
• Jupyter^® Notebook files are now displayed in a common format.

Speeding up development with SourceCraft CLI

The SourceCraft CLI with built-in agent mode allows you to manage repositories, tasks, change proposals, and other resources on the platform in command line mode.
What options are available:

• Performing multi-step tasks autonomously, from writing code to proposing changes in the background.
• Integration with development platforms already in use by teams.
• Data protection — the agent works on the basis of Yandex AI Studio language models, so code and work data are not transferred to external services.
• AI skills for running ready-made use cases.
• Easy authentication for Git without needing to explicitly use SSH keys or PAT.

SourceCraft course library

Together with the Center for Technology and Society, Yandex Cloud launched an open library of original courses on IT and AI, and we invite teachers and researchers to take part.

Serverless computing

Yandex Cloud Postbox

We implemented several updates to the service.

Storing email history with search and filtering

Allows users to view the detailed delivery history of each email, including statuses (sent, delivered, delayed), with the option to search and filter by sender, recipient, and subject.

This is critical for debugging newsletters and analyzing their effectiveness. The function is implemented without needing to store the contents of emails, ensuring privacy.

Processing user complaints

An internal mechanism for recording complaints about messages from recipients. The system processes notifications from email providers, updates sender reputation statistics, and can inform the client via the notification mechanism. Keeping track of the number of complaints and excluding such recipients from mass mailings is important for maintaining a good sender reputation and high deliverability.

Yandex Cloud Marketplace

April updates and new products in the Marketplace:

Akola

A no-code platform from RTK IT Plus for creating business applications. It makes it possible to develop internal systems, portals, and services via the visual designer: customize business objects, interfaces, and processes without programming. It is suitable for the rapid launch and development of digital products without its own infrastructure.

Wistium (from our partner SQEEL)

A platform for building corporate data platforms based on Data Mesh principles. It combines source connection, data profiling, ETL/ELT processes, modeling, and analytics production in a single loop. It supports a low-code approach and management of the full data lifecycle, from download to BI and ML scripts.

Open WebUI (with Yandex AI Studio integration)

A chat interface for working with language models and AI agents. The Marketplace build already includes integration with Yandex AI Studio: models do not need to be deployed on VMs; they are accessible via the cloud service. It is suitable for generating text, programming, and integrating AI into workflows.

Documentation