Aleph Alpha releases Kolibri, an open-weight model that makes the case for European AI sovereignty
Aleph Alpha has released Kolibri, an open-weight German-English language model with 78 billion parameters, of which about three billion are active per token through a mixture-of-experts architecture.
The company states that Kolibri sits on the Pareto front of quality and operating cost in both languages, outperforming compared models with similar architectures, some significantly older (March, April 2026), on either metric. According to the company, Kolibri scores 71% on German benchmarks while decoding faster than comparable models including GPT OSS A5B, Qwen 3.6 A3B, and Gemma 4 A4B.
German accounts for 21.3 percent of the training data, supported by a dedicated German data pipeline Aleph Alpha built for the project. Chinese models were also used to generate synthetic training data.
Aleph Alpha says Kolibri targets public administration, aviation, and industry. It was developed under European law with the EU AI Act in mind, supports context windows of up to one million tokens, and was trained on 768 B200 GPUs in Germany and Finland, according to the tech report. The weights are available under an Apache 2.0 license on Hugging Face.
Based on reporting from the original publisher. Visit the source for full context and later updates.
Publisher excerpt
Aleph Alpha has released Kolibri, a German-English mixture-of-experts model with 78 billion parameters, about three billion of which are active per token. Over 21 percent of the training data is German. The model was trained on 768 B200 GPUs in Germany and Finland, and the weights are freely available under the Apache 2.0 license. The article Aleph Alpha releases Kolibri, an open-weight model that makes the case for European AI sovereignty appeared first on The Decoder .