Germany’s Aleph Alpha releases Kolibri, a sovereign AI model for mission-critical government and industry applications

German AI firm Aleph Alpha has unveiled Kolibri, a specialised large language model (LLM) crafted for mission-critical government and industry uses. Launched on October 3, 2026, Aleph Alpha describes Kolibri as a sovereign AI model made in Germany, designed to meet the stringent demands of organisations that prioritise data control, documented compliance, and a model tailored to operate natively in German.

Ilhan Scheer, CEO of Aleph Alpha, stated, "Kolibri demonstrates that we possess the talent and expertise in Germany to craft competitive AI models. For us, AI sovereignty represents freedom of choice, retaining the capability to develop and advance this technology while allowing customers to control how they employ it. Our goal is to integrate this capability into everyday operations in both industry and public administration."

Founded in 2019, Aleph Alpha aims to create sovereign, human-centric AI. Mentioned alongside Mistral, the company stands as one of the two European entities challenging leading US and Chinese AI labs. Aleph Alpha employs approximately 200 individuals across four locations in Germany and utilises its specialised LLMs for enterprises and government institutions seeking data sovereignty and trustworthy AI applications.

Kolibri is an innovative mixture-of-experts (MoE) model boasting 78 billion total parameters, with roughly 3 billion active per token. The model is engineered so that during inference, it selectively activates only the necessary parts of the model, a design choice that boosts efficiency without sacrificing performance. Aleph Alpha notes that around 23% of Kolibri's pre-training data consists of German-language text, and the company has developed a German-optimised tokeniser that represents German text using fewer tokens, enabling faster and more efficient processing of German content.

Kolibri supports agentic workflows, retrieval-augmented generation (RAG), and native tool calling. For RAG applications, the model is trained to answer questions based on supplied documents and will refrain from responding when those documents lack sufficient evidence, thus reducing hallucinations in high-stakes scenarios. Users can also adjust the balance between reasoning depth and response speed via built-in reasoning controls.

To satisfy the compliance and provenance requirements of regulated industries and government institutions, Aleph Alpha has screened Kolibri against a blocklist of over 4.5 million URLs, drawing from sources like the European Commission’s Piracy Watch List. The company has also individually examined every third-party dataset for licensing terms, lawful sourcing, and adherence to opt-outs.

In a technical report, Aleph Alpha details how Kolibri addresses copyright, data protection, and EU AI Act requirements, with its licence terms and supporting documents accessible on Hugging Face. Samuel Weinbach, co-founder and co-Chief Research Officer of Aleph Alpha, said, "We designed Kolibri around the needs of German-language applications, from a tokeniser optimised for German to training the model to reason in German.

For our customers, this is a significant step towards making AI useful in everyday workflows." Following the launch of Kolibri, Aleph Alpha also announced a partnership with Canadian AI firm Cohere to establish a joint transatlantic sovereign AI company, subject to regulatory approval.