What happened
- Kolibri is a German-English language model with 78 billion parameters.
- Kolibri was trained on 768 B200 GPUs in Germany and Finland.
- Kolibri is available under an Apache 2.0 license on Hugging Face.
Why it matters
Kolibri's release highlights Aleph Alpha's commitment to European AI sovereignty by using German data and open-source licensing, positioning it as a competitive alternative in the global AI landscape.
The Elephant take
π ιΌ Kolibri is a bold move for European AI autonomy, but the focus on German data might limit its global appeal. The model's open-source nature is a strong statement, but will it outperform other models in practice?
Who should care
- AI Researchers
- European Governments
- Tech Companies
What to do next
- Evaluate Kolibri's performance in real-world applications
- Monitor the impact of European AI initiatives
- Explore open-source AI models for specific use cases
Keep in mind
The model's performance and real-world impact remain to be seen, and the emphasis on German data could raise questions about its global relevance.