Kolibri-1: Aleph Alpha's 78B sovereign open-weight MoE LLM
Aleph Alpha released Kolibri-1 on 3 Oct 2026 (German Unity Day): a 78B MoE (3.46B active per token) bilingual German-English model with up to 1M-token context, trained on infrastructure in Germany and Finland, weights on Hugging Face under Apache 2.0.
Why it matters
First big European sovereign open-weight model: bilingual-by-design German (not translated), 1M-token context, Apache 2.0 weights, runnable on infrastructure EU organizations control.
What you could build with it
A Berlin-based law firm could self-host Kolibri to summarize German legal documents in-house, keeping client data inside EU borders instead of sending it to a foreign API.
Does it hold up?
Too early to judge on real workloads: coverage is launch-day, though one independent teardown (tej.as) works through the 189-page technical report. Note the catch: 78 GB FP8 weights demand datacenter GPUs, so openness does not equal laptop-runability.
Built with Kolibri-1
- Kolibri is an open-weight LLM from Aleph Alpha for German and English — HN signallaxima · Hacker News thread hit 252 points and 115 comments within hours of the launch.
- Aleph Alpha Kolibri: Germany's Open-Weight AI Modelmadrobot · Launch-day writeup: 78B MoE, 3.5B active per token, 1M context, Apache 2.0 weights, timed for German Unity Day.
- Aleph Alpha opens a German AI model with a million-token contextalextech · Details the hardware bar: ~78 GB of weights (FP8) needing datacenter-class GPUs, plus the vLLM integration package.
Learn more
First spotted on article: source.
More AI models
ElevenLabs Eleven v4 + v4 Turbo: new emotive TTS architecture with a 100ms real-time variantElevenLabs released Eleven v4 and Eleven v4 Turbo on September 28: a new text-to-speech architecture with inline…models · JEV 0.72VoiceStudio: open-source, fully-local ElevenLabs alternative with voice cloning, dubbing and dictation in 646 languagesVoiceStudio is an open-source, fully-local voice AI studio: voice cloning from a 3-second sample, voice design, video…models · JEV 0.71DeepSeek V4-Flash official API: public beta with upgraded agent capabilities, Responses API and Codex supportDeepSeek's official V4-Flash is now a public beta API with massively upgraded agent capabilities - benchmark scores…models · JEV 0.69Mistral Forge: enterprise platform for training and continuously improving proprietary modelsMistral AI launched Forge (Sep 28) — a full-lifecycle model training platform (pre-training, SFT, DPO/ODPO, RL…models · JEV 0.69kev: open, trainable Jev-like family of small decision models on Qwen3.5/3.8A family of small decision models (0.8B to 27B) built on Qwen3.5/3.8 that reproduces Jev's prefill-only typed-decision…models · JEV 0.68GPT-Synopsys: OpenAI and Synopsys multi-year deal to build an AI model that operates EDA toolsOpenAI and Synopsys signed a multi-year agreement on Sep 30 to jointly develop GPT-Synopsys, a specialized model that…models · JEV 0.66
Get the week's best AI launches, plus 3 ideas worth building
One email every Saturday. Ranked by traction, not hype. Free.