Diabase/reranker-1
Text Classification • 0.2B • Updated • 27 • 1
AI & ML interests
None defined yet.
Recent Activity
Organization Card
Diabase
European AI infrastructure for sovereign, multilingual enterprise deployment.
We build open-weight AI models and tools designed for European organizations that need control over their data, compliance with European regulations, and strong multilingual support.
Models
| Model | Description | Hugging Face |
|---|---|---|
| Diabase Europa 9B v1.0 | Swedish-enhanced European assistant (Gemma 2 9B). Run locally with Ollama. | diabase/europa-9b-GGUF |
| Diabase Europa 9B LoRA | LoRA adapter for continued fine-tuning (advanced) | diabase/europa-9b-v1.0-lora |
| diabase-embedding-1 (research preview) | Swedish-first EuroBERT embedding model for semantic search and RAG | Diabase/embedding-1 |
| diabase-reranker-1 (research preview) | Swedish-first EuroBERT cross-encoder for reranking retrieved documents | Diabase/reranker-1 |
Quick start
huggingface-cli download diabase/europa-9b-GGUF gemma-2-9b-it.Q4_K_M.gguf Modelfile
ollama create diabase-europa-9b -f Modelfile
ollama run diabase-europa-9b
Principles
- Open weights, transparent evaluation — We publish benchmarks, limitations, and training approach.
- European data sovereignty — Models designed for local and on-prem deployment.
- Swedish-first — Dedicated tuning for natural Swedish assistant behavior.
- Honest claims — Early-stage models are labeled as such.
Links
License
Gemma-based model derivatives follow the Gemma Terms of Use. The EuroBERT-based retrieval models are released under Apache 2.0. Diabase artifacts and documentation are Apache 2.0 where applicable.
datasets 0
None public yet