Germany's Aleph Alpha Unveils Kolibri: A Sovereign AI Model for Mission-Critical Applications

Jan de Vries · 2026-10-05

Germany's Aleph Alpha has just released Kolibri, a new large language model (LLM) built specifically for mission-critical applications in government and industry. This isn't just another AI model—it's a statement that Europe can develop competitive, sovereign AI. ### What Makes Kolibri Special? Kolibri is a mixture-of-experts (MoE) model with 78 billion total parameters, but it only activates about 3 billion per token. That means it's incredibly efficient without sacrificing capability. Think of it like a team of specialists: you don't need everyone to answer every question, just the right expert for the task. The model is designed for organizations that need data control, documented compliance, and native German language support. Aleph Alpha says German text makes up about 23% of its pre-training data, and they've built a German-optimized tokenizer to process German faster and more accurately. ### Built for Trust and Compliance Aleph Alpha has gone to great lengths to ensure Kolibri meets the strict demands of regulated industries. They screened the model against a blocklist of over 4.5 million URLs, including the European Commission's Piracy Watch List. They also reviewed every third-party dataset for licensing and lawful sourcing. In a technical report, the company details how Kolibri addresses copyright, data protection, and EU AI Act requirements. The model's license terms and supporting documents are available on Hugging Face. ### Key Features for Real-World Use - **Agentic workflows**: Kolibri can handle complex, multi-step tasks. - **Retrieval-augmented generation (RAG)**: It answers questions based on provided documents and abstains when evidence is insufficient, reducing hallucinations. - **Native tool calling**: It integrates with other tools seamlessly. - **Reasoning controls**: Users can adjust the balance between reasoning depth and response speed. As Ilhan Scheer, CEO of Aleph Alpha, puts it: "Kolibri demonstrates that we have the talent and expertise in Germany to develop competitive AI models. For us, AI sovereignty means freedom of choice... and giving customers control over how they use it." ### The Bigger Picture: A Transatlantic Merger Kolibri's launch follows Aleph Alpha's April announcement of a merger with Canadian AI firm Cohere. The combined entity will operate globally as Cohere, with headquarters and R&D centers in both Canada and Germany. While the merger awaits regulatory approval, it signals a new era of transatlantic sovereign AI. With Kolibri now available and the merger in the pipeline, Aleph Alpha is proving that Europe can build its own large language models—without compromising on sovereignty. For organizations that need control, compliance, and top-tier performance, Kolibri is a compelling option.