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.

Get startedTalk to an expert

A wide range of models for different cases

Yandex’s cutting-edge developments and current open-source solutions.

Out-of-the-box inference and scaling

No need to purchase or rent hardware, DevOps teams, or complex operations.

Common interfaces and APIs

A unified way of working with models from different manufacturers, and OpenAI compatibility.

Enterprise data security

Access control, environment isolation, and enterprise compliance.

DeepSeek V4 Flash in Model Gallery

Plan actions, keep context in long chains of reasoning, and invoke tools without losing logic. It works confidently with code: generating, analyzing, and refactoring, understanding dependencies, and performing multi-step tasks.

Get started

A variety of models for your needs

Choose the right model for a specific use case, without experimentation or unnecessary complexity.

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

DeepSeek V4 Flash

Designed for complex RAG scenarios where consistent information processing and a structured approach to problem solving are important.

YandexGPT Pro

For complex RAG scenarios and summarization with high quality and stable context.

Qwen3.6 35B

Optimal for RAG scenarios that require high speed search and response generation.

Qwen3 235B

Effective for documentation search and summarization scenarios when working with large documents and multi-step analysis.

Yandex models

Open-source models

DeepSeek V4 Flash New

A model for code and agent scenarios and complex reasoning

Qwen3.6 35B

A multimodal model that works with text and images

Qwen3 235B

A model for agent scenarios and complex instructions

GPT OSS 20B и 120B

Language models for text generation, reasoning, and applied tasks

Test models right now

Playground is an interactive environment for experimenting with models. Select a model, enter your request, and see how the model responds in real time.
Models of all modalities are available: voice, images, text.

Go to Playground

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.

Responses API

Text generation and analysis

Realtime API

Streaming responses and voice scripts

Completions API

Working with language models, with easy integration in open-source frameworks

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.

Logging can be disabled.

Dedicated network accesses and connections between local infrastructure and the platform are available.

Learn more​

Pricing policy

Usage costs for 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.

Learn more

Options for accessing models

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

Instant responses

For interactive scenarios like chatbots, voice interfaces, and assistants. The models respond in real time, providing live interaction without delay.

Batch processing

For tasks with large amounts of data: mass text generation, document analysis, updates to knowledge bases. Requests are processed asynchronously and scalable in response to changing loads.

Dedicated inference

For deploying models outside a shared pool. It is suitable for specific scenarios and models not available on basic instances. Optimal for tasks requiring predictable resources and load management.

Other useful services

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

Documentation​

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

Documentation​

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.

Documentation​

Start working with Model Gallery

Try launching the first model — test responses in the console, connect the API, or train them according to the specifics of your business.

Get startedTalk to an expert