NbAiLab/borealis-270m-instruct-preview

TEXT GENERATIONPricing:Input $0.04 / Cached $0.008 / Output $0.08Concurrent Unit Cost:1Model Size:0.3BQuant:BF16Context Size:32kPublished:Jan 30, 2026License:gemmaArchitecture:Transformer0.0K Featherless Exclusive Cold

The NbAiLab/borealis-270m-instruct-preview is a 270 million parameter instruction-tuned preview model based on Google's Gemma 3 architecture, developed by NbAiLab. Fine-tuned exclusively on textual instructions, this experimental model is intended for early testing and feedback, focusing on Norwegian-centric assistant-style tasks. It is designed for assessing Norwegian writing style, quality, and language coverage (Bokmål/Nynorsk), despite being a pre-release quality model not yet safety-aligned.

Loading preview...

Model Overview

NbAiLab/borealis-270m-instruct-preview is a 270 million parameter instruction-tuned model, serving as a preview experiment from NbAiLab. It is built upon the Google Gemma 3 architecture and has been fine-tuned solely on textual instructions. This model is released for early testing and feedback, and users should be aware it is of pre-release quality.

Key Characteristics

  • Experimental Nature: This is a preview model intended for early evaluation and is not yet production-ready.
  • Instruction-Tuned: Fine-tuned on textual instructions using the NbAiLab/aurora-sft-2512 dataset.
  • Norwegian Focus: Specifically designed for tasks related to the Norwegian language, including Bokmål and Nynorsk.
  • Format Availability: Provided in various formats including original Transformers, GGUF quantizations (q8_0, f16, bf16), and MLX for Apple Silicon (32-bit and 8-bit).

Intended Use Cases

  • Norwegian Assistant Tasks: Suitable for drafting, summarization, Q&A, and light reasoning in a Norwegian context.
  • Writing Assessment: Useful for evaluating Norwegian writing style and quality.
  • Early Evaluation: Ideal for assessing model behavior, language coverage, and overall quality in a controlled environment.

Important Considerations

  • Safety Disclaimer: This model is not safety-aligned and may produce harmful, biased, or inappropriate content. It is not recommended for safety-critical applications without additional mitigations.
  • Preview Quality: Outputs may be unstable and prone to hallucinations due to its experimental status.

Finetunes
4 models