In just three months, the Kandinsky 2.1 generative model from Sber gave birth to over 70 million distinct images. The most searched themes among users included Russia, love, anime, cats, and space. This milestone was shared by Alexander Vedyakhin, the first deputy chairman of Sberbank’s board, highlighting how swiftly the technology is resonating with people across the country.
Vedyakhin stressed that the intense curiosity around neural networks from Russian users reflects a natural stage of growth for the nation, which stands at the forefront of artificial intelligence development. The momentum mirrors a broader trend of creative experimentation and rapid adoption of AI-powered tools at scale, propelled by a strong local ecosystem of researchers and developers.
He added that Kandinsky was created and continues to be developed to spark new creative ideas in individuals and to be freely accessible for turning those ideas into visual reality. The goal is simple: empower creators to explore possibilities without barriers, using a tool that adapts to their imaginations.
Earlier this year, Sber rolled out Kandinsky 2.1 as a productive model in early spring. Within four days of its launch, it had attracted one million users, marking it as one of the fastest-growing neural networks ever released to the public. This rapid adoption demonstrates how quickly AI-assisted creativity can scale when a platform offers powerful capabilities in a user-friendly package.
The Kandinsky 2.1 system was built by a team of researchers at Sber AI, with collaborative support from scientists at the AIRI Institute of Artificial Intelligence, and was trained on a combined dataset drawn from Sber AI resources and SberDevices. This collaboration underscores the value of pooling expertise and data to enhance the model’s versatility and performance across different creative tasks.
When given a natural language prompt, Kandinsky 2.1 can produce images in mere seconds. A user can select from twenty distinct visual styles, ranging from classical approaches to culturally specific textures and motifs, such as classical painting, mosaic patterns, Khokhloma decor, and the evocative brushwork of Aivazovsky, among others. This variety empowers creators to experiment with mood, technique, and heritage in a single search, making the tool suitable for a wide array of projects and audiences.
The model understands requests in both Russian and English, enabling a broad user base to craft pictures using an endless canvas mode. It also supports interactive editing features, including the ability to draw parts of an image and blend multiple drawings into a single composition. These capabilities expand the workflow for artists, designers, and educators who seek to iterate quickly and push creative boundaries—without switching tools or platforms. This flexibility is particularly valuable for teams working on multimedia presentations, concept art, and educational content, where rapid visualization can accelerate decision-making and learning outcomes.
Industry observers and stakeholders have discussed the maturation of AI in the creative sector, noting how tools like Kandinsky 2.1 can redefine workflows, enable new forms of collaboration, and unlock opportunities across publishing, advertising, game development, and digital art. The ongoing evolution of such models is shaping how artists approach idea generation, iteration, and final production, with implications for both individual creators and larger organizations. Analysts emphasize the importance of responsible use, clear attribution, and thoughtful moderation to ensure AI-generated imagery respects copyright and cultural sensitivities while still enabling bold experimentation. In this light, Kandinsky 2.1 represents a notable step in the ongoing dialogue about AI-assisted creativity and its role in contemporary visual culture.