Aleph Alpha launches Kolibri sovereign AI model

Heidelberg-based AI company Aleph Alpha has released Kolibri, a new bilingual (German and English) language model designed for public administration and industrial use. The model utilizes a mixture-of-experts architecture with 78 billion total parameters, of which 3 billion are active per token.

Kolibri is specifically engineered for sovereignty, allowing customers to run the model on their own controlled infrastructure to ensure data privacy and control. The model features a context window of up to one million tokens and includes four reasoning levels. Aleph Alpha has made the model weights available under the Apache 2.0 open-source license on Hugging Face.

The release follows Aleph Alpha’s announced transatlantic merger with the Canadian AI firm Cohere, which functions as an acquisition to develop sovereign AI offerings for sectors including finance, defense, and healthcare.