NbAiLab/borealis-12b

Borealis 12B: Norwegian-Centric Instruction-Tuned Model

NbAiLab/borealis-12b is a 12-billion parameter instruction-tuned model developed by the National Library of Norway (NbAiLab). It is part of the Borealis family, built upon the Gemma 3 architecture (specifically, fine-tuned from google/gemma-3-12b-it). This model is a "full release" version, meaning it incorporates a limited subset of data from an agreement with Norwegian rights-holder organizations, primarily to enhance its ability to generate news article titles and ingress texts.

Key Capabilities

• Norwegian-centric Instruction Following: Optimized for tasks requiring understanding and generation in Norwegian.
• Assistant-Style Tasks: Proficient in drafting, summarization, and question answering.
• Specialized Content Generation: Incorporates unique training data for generating news article titles and ingress texts.
• Safety Alignment: Utilizes "prompt baking" to distill desired behaviors directly into model weights, balancing quality and safety.

Good For

• Norwegian Language Applications: Ideal for developers building applications focused on the Norwegian language, including Bokmål and Nynorsk.
• Content Creation: Useful for generating various forms of Norwegian text, particularly news-related content.
• Research and Evaluation: Suitable for assessing and improving Norwegian writing style and language coverage.
• Fine-tuning and Customization: Provides a strong base for further fine-tuning on specific Norwegian datasets or tasks.