Scenarios for search, analysis, and summarizing information based on corporate data. Used in assistants, internal portals, analytics, and document management.

Model Gallery
A catalog of models for any task: from generating text and images to working with unstructured data. Here are models for working with text, images, and voice.
A wide range of models for different scenarios
Yandex’s cutting-edge developments and current open-source solutions.
Out-of-the-box inference and scaling
Without purchasing or renting hardware, DevOps teams, or complex operations.
Common interfaces and APIs
One way of working with models from different manufacturers, OpenAI compatibility.
Enterprise data security
Access control, environment isolation, and Enterprise compliance.
DeepSeek V3.2 in Model Gallery
Plans actions, keeps context in long chains of reasoning, and invokes tools without losing logic. It works confidently with code: generating, analyzing and refactoring, understanding dependencies and performing multi-step tasks.

A variety of models for your needs
Choose the right model for a specific use case, without experimentation or unnecessary complexity.

Yandex models
Opensource models

DeepSeek V3.2 New

Gemma 3 27B

Qwen3 235B

GPT OSS 20B и 120B
Test the models right now
Playground is an interactive environment for experimenting with models. Select a model, enter your request, and see how it responds in real time.
Models of all modalities are available: voice, images, text.

Common interfaces and standards
Use familiar SDKs and frameworks — LangChain, LlamaIndex, LangGraph — with minimal code changes. Model Gallery is compatible with the OpenAI interface and supports core developer tools.
Secure work with corporate data
Model Gallery is focused on the safe and predictable use of models, rather than on fine-tuning on customer data.
Data is not used to train models.
Accesses are controlled at the platform level.
Option to disable logging.
Dedicated network accesses and connections between local infrastructure and the platform.

Pricing policy
The cost of using models from Model Gallery depends on the model’s operating mode, the number of incoming and outgoing tokens, and the tools used. The token count in the same text may vary from one model to another.

Options for accessing models
Model Gallery supports various inference modes, for tasks of any complexity and volume.

Instant responses

Batch processing

Dedicated inference
Other useful services

A service for receiving Yandex search database responses in XML format or HTML. It helps organize search on a site, a group of sites, or the internet, and track the position of sites for search queries.

Computer vision service for recognizing text in images and PDF files. It supports 45+ languages and detects them automatically.

A service for integrating Yandex Translator algorithms into applications or web projects for end users. It supports 100+ languages and can translate individual words and entire texts.
Start working with Model Gallery
Try launching the first model — test the responses in the console, connect the API, or train them according to the specifics of your business.

