Aleph Alpha Releases 78.1B-Parameter Kolibri Model
Aleph Alpha has released Kolibri-1, an open-weight Mixture-of-Experts language model designed for English and German. The model contains 78.1 billion parameters, but activates 3.46 billion, or 4.4%, for each token. It supports contexts of up to 1,048,576 tokens and allows reasoning effort to be selected for individual requests. Its weights are available under the Apache 2.0 license on Hugging Face.
The FP8 checkpoint is about 78GB and can run on one B200, B300 or H200 GPU, or on two H100 SXM5 GPUs. Deployment is supported through vLLM with dedicated reasoning and tool-call parsers. Kolibri was developed end to end by teams in Germany, with training infrastructure in Germany and Finland. Aleph Alpha says its design addresses requirements associated with the EU AI Act, GDPR and the EU General-Purpose AI Code of Practice.
Training covered about 24 trillion tokens, including more than 2 trillion curated or synthetically generated German tokens. In the company’s evaluations, Kolibri scored 75.5 in English and 70.8 in German overall, led GPQA Diamond and AIME results in English, and achieved an 89.3 English code score. Its agentic average was 63.4 in English, while its BFCL v4 score was 61.4.
Written by Sakhanda Wire AI from reporting by MarkTechPost. Facts, quotes and full context are in the original article below.