borealis-12B-instruct-preview: benchmark results
NB-AiLab at the National Library of Norway publishes Borealis 12B Instruct Preview, a Norwegian instruction fine-tune of google/gemma-3-12b-it. This is the experimental Transformers checkpoint fine-tuned on textual instructions, distinct from its GGUF and MLX conversion releases. Provider: National Library of Norway. Released 2026-01-31. Access: Open.
Aggregate Rating 1590 ± 34, rank #1125 of 10916 rated models, from 47 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| EuroEval Swedish NLU - Swerec | 80.43 | Matthews correlation coefficient (%) | 97.1 |
| EuroEval Norwegian NLU - Norec | 59.85 | Matthews correlation coefficient (%) | 90.4 |
| EuroEval Faroese NLU - FoSent | 61.8 | Matthews correlation coefficient (%) | 88.5 |
| EuroEval Faroese | 55.9 | Average Score (%) | 87.5 |
| EuroEval Icelandic NLU - Hotter and Colder Sentiment | 52.52 | Matthews correlation coefficient (%) | 87.5 |
| EuroEval Danish NLU - Angry Tweets | 56.3 | Matthews correlation coefficient (%) | 87.1 |
| EuroEval Faroese NLU - FoQA | 74.42 | Character F1 (%) | 86.7 |
| EuroEval Danish NLU | 59.41 | NLU Average Score (%) | 82.1 |
| EuroEval Danish Common Sense Reasoning | 67.42 | Common Sense Reasoning Average Score (%) | 81.7 |
| EuroEval Icelandic NLU - MIM-GOLD NER | 65.12 | Micro F1 excluding MISC (%) | 81.3 |
| EuroEval Norwegian NLU | 62.26 | NLU Average Score (%) | 81.1 |
| EuroEval Norwegian Common Sense Reasoning | 77.46 | Common Sense Reasoning Average Score (%) | 80.1 |
Interactive version: theaggregate.ai/model?slug=borealis-12b-instruct-preview · How It Works · Data refreshed daily, snapshot 2026-11-08.