File size: 3,386 Bytes
6768c29 1c3d460 1c07917 6768c29 b690ea2 6d91fc9 cba1721 b690ea2 6554992 b690ea2 3edd095 b690ea2 546e802 b690ea2 3edd095 b690ea2 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 |
---
license: gemma
datasets:
- NbAiLab/aurora-sft-2512-filtered
language:
- 'no'
- nb
- nn
base_model:
- google/gemma-3-4b-it
pipeline_tag: image-text-to-text
library_name: transformers
tags:
- conversational
- instruct
- experimental
---
# 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**](https://huggingface.co/google/gemma-3-4b-it), and fine-tuned on textual instructions only.
| Model | Bits | Format |
|---|---:|---|
| [NbAiLab/borealis-4b-instruct-preview](https://huggingface.co/NbAiLab/borealis-4b-instruct-preview) | BF16 | Transformers (safetensors) |
| [NbAiLab/borealis-4b-instruct-preview-gguf](https://huggingface.co/NbAiLab/borealis-4b-instruct-preview-gguf) | 8 | GGUF (`q8_0`) |
| [NbAiLab/borealis-4b-instruct-preview-gguf](https://huggingface.co/NbAiLab/borealis-4b-instruct-preview-gguf) | 16 | GGUF (`f16`) |
| [NbAiLab/borealis-4b-instruct-preview-gguf](https://huggingface.co/NbAiLab/borealis-4b-instruct-preview-gguf) | BF16 | GGUF (`bf16`) |
| [NbAiLab/borealis-4b-instruct-preview-mlx](https://huggingface.co/NbAiLab/borealis-4b-instruct-preview-mlx) | 32 | MLX |
| [NbAiLab/borealis-4b-instruct-preview-mlx-8bits](https://huggingface.co/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**](https://huggingface.co/NbAiLab/borealis-4b-instruct-preview-gguf):
- `model-q8_0.gguf`
- `model-f16.gguf`
- `model-bf16.gguf`
Use:
```bash
ollama run hf.co/NbAiLab/borealis-4b-instruct-preview-gguf:BF16
```
### MLX (Apple Silicon)
Available in [**NbAiLab/borealis-4b-instruct-preview-mlx**](https://huggingface.co/NbAiLab/borealis-4b-instruct-preview-mlx) and quantized to [8 bits](https://huggingface.co/NbAiLab/borealis-4b-instruct-preview-mlx-8bits).
Use:
```bash
# 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. |