81 lines
4.9 KiB
Markdown
81 lines
4.9 KiB
Markdown
---
|
|
license: gemma
|
|
datasets:
|
|
- NbAiLab/aurora-sft-2512-filtered
|
|
language:
|
|
- 'no'
|
|
- nb
|
|
- nn
|
|
base_model:
|
|
- google/gemma-3-270m-it
|
|
pipeline_tag: text-generation
|
|
library_name: transformers
|
|
tags:
|
|
- conversational
|
|
- instruct
|
|
- experimental
|
|
---
|
|
|
|
# Borealis 270M Instruct (Preview)
|
|
|
|
Release: Jan 31st, 2026.
|
|
|
|
## Model summary
|
|
**NbAiLab/borealis-270m-instruct-preview** is a **270M-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-3-270m-it**](https://huggingface.co/google/gemma-3-270m-it), and fine-tuned on textual instructions only.
|
|
|
|
| Size | Transformers | GGUF (quants: `q8_0`, `f16`, `bf16`) | MLX |
|
|
|---:|---|---|---|
|
|
| 270M | [NbAiLab/borealis-270m-instruct-preview](https://huggingface.co/NbAiLab/borealis-270m-instruct-preview) | [NbAiLab/borealis-270m-instruct-preview-gguf](https://huggingface.co/NbAiLab/borealis-270m-instruct-preview-gguf) | [NbAiLab/borealis-270m-instruct-preview-mlx](https://huggingface.co/NbAiLab/borealis-270m-instruct-preview-mlx) (32-bit)<br>[NbAiLab/borealis-270m-instruct-preview-mlx-8bits](https://huggingface.co/NbAiLab/borealis-270m-instruct-preview-mlx-8bits) (8-bit) |
|
|
| 1B | [NbAiLab/borealis-1b-instruct-preview](https://huggingface.co/NbAiLab/borealis-1b-instruct-preview) | [NbAiLab/borealis-1b-instruct-preview-gguf](https://huggingface.co/NbAiLab/borealis-1b-instruct-preview-gguf) | [NbAiLab/borealis-1b-instruct-preview-mlx](https://huggingface.co/NbAiLab/borealis-1b-instruct-preview-mlx) (32-bit)<br>[NbAiLab/borealis-1b-instruct-preview-mlx-8bits](https://huggingface.co/NbAiLab/borealis-1b-instruct-preview-mlx-8bits) (8-bit) |
|
|
| 4B | [NbAiLab/borealis-4b-instruct-preview](https://huggingface.co/NbAiLab/borealis-4b-instruct-preview) | [NbAiLab/borealis-4b-instruct-preview-gguf](https://huggingface.co/NbAiLab/borealis-4b-instruct-preview-gguf) | [NbAiLab/borealis-4b-instruct-preview-mlx](https://huggingface.co/NbAiLab/borealis-4b-instruct-preview-mlx) (32-bit)<br>[NbAiLab/borealis-4b-instruct-preview-mlx-8bits](https://huggingface.co/NbAiLab/borealis-4b-instruct-preview-mlx-8bits) (8-bit) |
|
|
| 12B | [NbAiLab/borealis-12b-instruct-preview](https://huggingface.co/NbAiLab/borealis-12b-instruct-preview) | [NbAiLab/borealis-12b-instruct-preview-gguf](https://huggingface.co/NbAiLab/borealis-12b-instruct-preview-gguf) | [NbAiLab/borealis-12b-instruct-preview-mlx](https://huggingface.co/NbAiLab/borealis-12b-instruct-preview-mlx) (32-bit)<br>[NbAiLab/borealis-12b-instruct-preview-mlx-8bits](https://huggingface.co/NbAiLab/borealis-12b-instruct-preview-mlx-8bits) (8-bit) |
|
|
| 27B | [NbAiLab/borealis-27b-instruct-preview](https://huggingface.co/NbAiLab/borealis-27b-instruct-preview) | [NbAiLab/borealis-27b-instruct-preview-gguf](https://huggingface.co/NbAiLab/borealis-27b-instruct-preview-gguf) | [NbAiLab/borealis-27b-instruct-preview-mlx](https://huggingface.co/NbAiLab/borealis-27b-instruct-preview-mlx) (32-bit)<br>[NbAiLab/borealis-27b-instruct-preview-mlx-8bits](https://huggingface.co/NbAiLab/borealis-27b-instruct-preview-mlx-8bits) (8-bit) |
|
|
|
|
|
|
## 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-270m-instruct-preview** (safetensors).
|
|
|
|
### GGUF quantizations
|
|
Available in [**NbAiLab/borealis-270m-instruct-preview-gguf**](https://huggingface.co/NbAiLab/borealis-270m-instruct-preview-gguf):
|
|
- `model-q8_0.gguf`
|
|
- `model-f16.gguf`
|
|
- `model-bf16.gguf`
|
|
|
|
Use:
|
|
```bash
|
|
ollama run hf.co/NbAiLab/borealis-270m-instruct-preview-gguf:BF16
|
|
```
|
|
|
|
### MLX (Apple Silicon)
|
|
Available in [**NbAiLab/borealis-270m-instruct-preview-mlx**](https://huggingface.co/NbAiLab/borealis-270m-instruct-preview-mlx) and quantized to [8 bits](https://huggingface.co/NbAiLab/borealis-270m-instruct-preview-mlx-8bits).
|
|
|
|
Use:
|
|
```bash
|
|
# Install MLX LM
|
|
uv tool install mlx-lm
|
|
|
|
# Interactive chat REPL
|
|
mlx_lm.chat --model "NbAiLab/borealis-270m-instruct-preview-mlx"
|
|
```
|
|
|
|
## Acknowledgements
|
|
Thanks to the **Gemma** team at Google for releasing Gemma 3 and to everyone contributing feedback on this preview. |