NbAiLab/borealis-4b-instruct-preview

Borealis 4B Instruct (Preview)

Release: Dec 22nd, 2025.

Model summary

NbAiLab/borealis-4b-instruct-preview is a 4B-parameter instruction-tuned preview model intended for early testing and feedback. It is an experiment and should be treated as pre-release quality.

This model is based on google/gemma-3-4b-it, and fine-tuned on textual instructions only.

Model	Bits	Format
NbAiLab/borealis-4b-instruct-preview	BF16	Transformers (safetensors)
NbAiLab/borealis-4b-instruct-preview-gguf	8	GGUF (q8_0)
NbAiLab/borealis-4b-instruct-preview-gguf	16	GGUF (f16)
NbAiLab/borealis-4b-instruct-preview-gguf	BF16	GGUF (bf16)
NbAiLab/borealis-4b-instruct-preview-mlx	32	MLX
NbAiLab/borealis-4b-instruct-preview-mlx-8bits	8	MLX (quantized)

Training data

Supervised fine-tuning (SFT) uses NbAiLab/aurora-sft-2512 (not released yet).

⚠️ Safety / alignment disclaimer (important)

This is a preview experiment and has not been safety-aligned yet. The model may produce harmful, biased, or insensitive outputs (including content that is offensive, unsafe, or inappropriate). Do not use it for safety-critical or high-stakes applications, and add your own safety mitigations if deploying.

Intended use

• Norwegian-centric assistant-style tasks (e.g., drafting, summarization, Q&A, light reasoning).
• Assesstment of Norwegian writing style and quality.
• Early evaluation of behavior, language coverage (Norwegian / Bokmål / Nynorsk), and quality.

Limitations

• Preview quality; outputs may be unstable and may hallucinate.
• Not aligned for safety; may follow harmful instructions or generate problematic content (see disclaimer above).

Weights & formats

Transformers (original)

• NbAiLab/borealis-4b-instruct-preview (safetensors).

GGUF quantizations

Available in NbAiLab/borealis-4b-instruct-preview-gguf:

• model-q8_0.gguf
• model-f16.gguf
• model-bf16.gguf

Use:

ollama run hf.co/NbAiLab/borealis-4b-instruct-preview-gguf:BF16

MLX (Apple Silicon)

Available in NbAiLab/borealis-4b-instruct-preview-mlx and quantized to 8 bits.

Use:

# Install MLX LM
uv tool install mlx-lm

# Interactive chat REPL
mlx_lm.chat --model "NbAiLab/borealis-4b-instruct-preview-mlx"

Acknowledgements

Thanks to the Gemma team at Google for releasing Gemma 3 and to everyone contributing feedback on this preview.