galactica-30B: benchmark results
Meta AI Papers with Code Galactica 30B is a scientific decoder-only language model trained on a 106B-token scientific corpus, including papers, textbooks and structured scientific data. This is the original base release under CC BY-NC 4.0, separate from OPT and instruction-tuned variants. Provider: Other. Access: Open.
Aggregate Rating 1273 ± 19, rank #9031 of 10916 rated models, from 68 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| MMLU-by-task - Electrical Engineering | 56.55 | Accuracy (%) | 91.6 |
| MMLU-by-task - Anatomy | 54.81 | Accuracy (%) | 90.7 |
| MMLU-by-task - High School Physics | 37.75 | Accuracy (%) | 89.6 |
| MMLU-by-task - College Physics | 34.31 | Accuracy (%) | 86.4 |
| MMLU-by-task - High School Chemistry | 44.83 | Accuracy (%) | 84.1 |
| MMLU-by-task - Econometrics | 34.21 | Accuracy (%) | 82.6 |
| MMLU-by-task - Conceptual Physics | 47.23 | Accuracy (%) | 82.2 |
| MMLU-by-task - College Mathematics | 36 | Accuracy (%) | 80.2 |
| MMLU-by-task - College Medicine | 50.29 | Accuracy (%) | 73 |
| MMLU-by-task - College Biology | 58.33 | Accuracy (%) | 72.8 |
| MMLU-by-task - Medical Genetics | 58 | Accuracy (%) | 71.9 |
| MMLU-by-task - College Computer Science | 45 | Accuracy (%) | 71.3 |
Interactive version: theaggregate.ai/model?slug=galactica-30b · How It Works · Data refreshed daily, snapshot 2026-11-08.