Show HN: Germany's new sovereign AI model Kolibri
Aleph Alpha released Kolibri, an Apache 2.0 78B MoE model for German and English.
Aleph Alpha released Kolibri 1 on 3 October 2026, an Apache 2.0 open-weight mixture-of-experts model for German and English. It has 78.1 billion parameters with about 3.46 billion active per token, a native 262,144-token context tested to 1,048,576, and a 18 June 2026 knowledge cutoff. Training used roughly 24 trillion tokens, more than a fifth German, on 768 NVIDIA B200 GPUs in Germany and Finland. Aleph Alpha says it was built with the EU AI Act in mind and signed the EU GPAI Code of Practice; some data was rephrased with Gemma 4 and Mistral-NeMo, and Qwen3-32B labeled quality filters.
- 78.1B parameters with 3.46B active per token
- Apache 2.0 weights; about 24 trillion training tokens
- Native 262,144-token context, tested to 1,048,576
- German tokenizer uses 11.2% fewer tokens than GPT-5
- Trained in Germany and Finland on 768 NVIDIA B200s
- Vendors
- Aleph AlphaNVIDIA
- Products
- Kolibri
- Organizations
- Aleph AlphaEuropean Union
- AI models
- KolibriGemma 4Mistral-NeMoQwen3-32B
From the original article
12 min read at tej.as
Kolibri is an open-weight large language model (LLM) from Aleph Alpha for German and English: a mixture of experts with 78 billion parameters that only uses about 3. 5 billion of them for each token it reads or writes. It came out on 3 October 2026 under the Apache 2.
By tejaskumar__ · tej.as
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.