NbAiLab/borealis-27b-instruct-preview
NbAiLab/borealis-27b-instruct-preview is a 27-billion parameter instruction-tuned preview model developed by NbAiLab, based on Google's Gemma-3-27b-it architecture. This experimental model is fine-tuned exclusively on textual instructions, specifically for Norwegian-centric assistant-style tasks such as drafting, summarization, Q&A, and light reasoning. It is intended for early testing and feedback, with a notable context length of 32768 tokens, and is not yet safety-aligned.
Loading preview...
Borealis 27B Instruct (Preview)
NbAiLab/borealis-27b-instruct-preview is a 27-billion parameter instruction-tuned model from NbAiLab, released as a preview for early testing and feedback. This experimental model is based on the google/gemma-3-27b-it architecture and has been fine-tuned solely on textual instructions using the NbAiLab/aurora-sft-2512 dataset.
Key Characteristics
- Preview Quality: This is an experimental, pre-release model, meaning outputs may be unstable and prone to hallucinations.
- Norwegian Focus: Specifically fine-tuned for Norwegian-centric assistant tasks, including drafting, summarization, Q&A, and light reasoning.
- Safety Disclaimer: The model has not been safety-aligned and may produce harmful, biased, or insensitive content. It is not suitable for safety-critical applications without additional mitigations.
Good for
- Early evaluation of model behavior and language coverage for Norwegian (Bokmål / Nynorsk).
- Assessment of Norwegian writing style and quality.
- Non-safety-critical assistant-style tasks in Norwegian contexts.
- Developers interested in providing feedback on an experimental instruction-tuned model.