Improved German versions of widely used LLM benchmarks
Boldt
university
AI & ML interests
None defined yet.
Recent Activity
Papers
View all Papers Organization Card
Welcome to Boldt!
Boldt is a family of German language models developed by the Chair of Machine Learning @ Humboldt-Universität zu Berlin. This organization hosts our models, datasets, and research artifacts related to the Boldt project.
Feel free to explore, download, and experiment with our latest releases! 🚀
🌟 The Boldt Model Family
Our models are trained on our German Dense-Core subset of FineWeb-2, utilizing a multi-epoch training recipe on high-quality data.
| Model | Parameters | Context Window | Description |
|---|---|---|---|
| Boldt-DC-350M | 350M | 2048 | Ultra-lightweight base model for constrained environments. |
| Boldt-DC-1B | 1B | 2048 | Highly optimized 1B base model with top-tier German performance. |
| Boldt-1B | 1B | 4096 | Extended context and augmented with 6B tokens of high-quality German news data. |
| Boldt-1B-IT-Preview | 1B | 4096 | Experimental instruction-tuned model. |
📊 Comparison
Boldt-1B compares favorably on German LLM benchmarks against other similarly-sized models:
It is even competitive with many larger (2B parameter) models. See our paper for the full evaluation.
📖 Research & Artifacts
datasets 8
Boldt/fineweb2-de-dc
Viewer • Updated • 24.4M • 24
Boldt/fineweb2-de-edu
Viewer • Updated • 30.2M • 33
Boldt/fineweb2-de-iv
Viewer • Updated • 43.4M • 38
Boldt/fineweb2-de-coherence
Viewer • Updated • 138M • 29
Boldt/obqa_de
Viewer • Updated • 495 • 22
Boldt/hellaswag_de
Viewer • Updated • 10k • 1.2k
Boldt/arc_de
Viewer • Updated • 3.44k • 1.25k
Boldt/lambada_openai_de
Viewer • Updated • 5.15k • 1.2k