Yandex releases open-source language model Alice AI Foundation LLM
Yandex has developed the large language model Alice AI Foundation LLM from scratch and opened its weights for commercial and research use, company representatives said.
The model's architecture features 80 billion parameters, with about 3 billion active at any one time. The training dataset volume amounted to 18 trillion tokens. The company clarified that training was conducted without using foreign third-party components, describing the development as fully sovereign, Vedomosti reported, citing a Yandex representative.
The developers released the tuned model parameters, enabling third-party specialists to deploy it on their own infrastructure and fine-tune it. The license does not restrict the model's application in commercial or research projects.
Yandex plans to use Alice AI Foundation LLM as a technological foundation for a future model with reasoning capabilities and to expand the agentic features of the Alisa AI assistant. Experts note that the current version is a base architecture rather than a finished product for end users.
What it means
Developers can freely download the open parameters of Alice AI Foundation LLM, deploy the model, and fine-tune it for commercial and scientific tasks.
Sources
- iXBT.com: Яндекс с нуля создал языковую модель на 80 млрд параметров media outlet
- Hi-Tech Mail: «Яндекс» разработал ИИ-модель без иностранных составляющих media outlet
This article was prepared with the help of AI and translated from the Russian original. Spotted an error — let us know.
Published 21 September 2026, 12:16 · No updates · Russian original
Discuss the news in our Telegram channel. Subscribe