Create vivid content for advertising campaigns and social networks. Experiment with different images and short texts to attract more customers to your site.
YandexART request
Cake, sign 'HAPPY', decorations of flowers and balloons on the background
Foundation Models
Foundation Models
We are making it easier to solve business tasks by opening up access to our foundation models. Integrate YandexGPT, YandexART, and other models into your applications, and we will develop GenAI together.
YandexGPT 5
The fifth generation of the Yandex generative language model is available to all users. Companies use YandexGPT API for text analysis, content creation, training chatbots, and solving other business tasks.
YandexART 2.0
This Yandex neural network generates images based on text description. You can use it to create visual content for advertising, presentations, websites, and apps.
Embeddings
YandexART provides an API to conveniently work with embeddings, i.e., vector text representation. Classify information, compare and align texts, or run a semantic search through your own knowledge base.
ClassifiersPreview
Recognize and categorize different types of objects using classifiers. You can train the classifier based on YandexGPT in DataSphere. This will allow you, for example, to speed up the distribution of support requests or filter out inappropriate content.
Fine-tuning
Fine-tune YandexGPT for a certain response format or text analysis rules, so the model more accurately grasps the specifics of your tasks, e.g., to analyze CVs and identify suitable candidates according to provided criteria, assess customers’ emotions in their reviews.
Creating AI assistants
The AI Assistant API allows you to easily and efficiently integrate large language models into complex, modular, and scalable applications by providing tools for managing context, memory, and external data (RAG).
How it works
Our foundation models are based on advanced technology and reliable Yandex Cloud infrastructure.
We provide ready-to-use models and regularly fine-tune LLM on new datasets to improve generation quality.
The new generation of YandexART
Our diffusion generative model will help you bring your ideas to life.
The new generation of the YandexART 2.0 neural network can generate short texts directly in an image and use several styles within one image, while better understanding the physics of space and creating more detailed images.
Realize and visualize customer ideas. In the request, you can set the level of detail, material type, and specify the color palette.
YandexART request
Light minimalist clean interior, lots of sunlight, abstract colourful painting on the wall, very high ceilings, sofa, coffee table, table lamp, white and milky, natural materials, realism, large space, soft pillows on the sofa
Create collections of branded products with elements of your corporate identity and using the relevant color scheme.
YandexART request
A loose-fitting hoodie in indigo color with a logo illustration in the center in the form of a white comet, female model. Minimalistic product photography on a light, monochromatic background
Bring creative ideas to live to promote a brand or company: come up with characters, design branded icons and logos.
YandexART request
Capybara, 3D animation style, on a plain yellow background holding a large slice of pizza
Create attractive illustrations of product categories for any media: website, social networks, printed materials.
YandexART request
Crispy baguette, hyper detail, focus on dish, bright lighting, aesthetically pleasing, beautiful, high resolution, super detail, 85mm lens, 1/250 sec, realistic, f/2.8, ISO 100, 4k, professional photography, close-up
What our models can currently do
We continue working on how to solve these tasks more efficiently with you.
Text generation and summary
LLM can create content based on a text prompt, highlight key ideas from a dataset, and visualize them in a concise and clear way. This saves considerable time and helps you automate routine tasks.
Semantic search
Foundation models can recognize prompt context and report more accurate results that meet user expectations, even if the search keywords do not match exactly.
Retrieval-augmented generation (RAG)
LLM can respond to prompts and generate texts considering external data sources. To do this, the language model trained on a large amount of data gets knowledge bases compiled specially for your domain.
Image creating
Generation is based on the cascade diffusion method. For model training, we change images by applying random noise. The task of the model is to restore these images. First, a color spot appears from the noise, and then the model gradually adds details.
Tryout Foundation Models in demo modes
Why choose Yandex Cloud
Get all the resources you need to train and deploy models from the start. Easily scale your projects, if required, without purchasing additional capacity.
What we offer
API-based integration
This might come in handy if you want to automate your company workflows. To use our API, log in and make sure your billing account is connected to the cloud.
Playground
Use Playground, a Yandex Cloud management console interface, to learn about service features and test model responses. We’ve got various demo modes to choose from.
Our partners
FAQ
This service is useful for everyone looking to accelerate business development with ML technologies: product managers, developers, fintech CDOs, as well as retail, industry, and experts in digital service development.