GALAI is the Python API for GALACTICA, a family of large language models trained on scientific text and data by Meta AI and Papers With Code, published on GitHub with about 2,700 stars. It makes the models installable from pip and usable in a few lines of code.
GALACTICA was trained on a large corpus of papers, textbooks, code, equations and structured scientific knowledge, and is designed for scientific NLP tasks: it can draft academic text, predict citations, reason through mathematics, annotate proteins and predict molecular properties. The galai package wraps the released checkpoints, which range from 125 million to 120 billion parameters, behind a simple generation API with sensible defaults for scientific prompts and special tokens for citations, LaTeX and SMILES strings. The project is Apache-2.0 licensed; the codebase dates from the model’s 2023 open release period. Example prompts for tasks such as solving textbook problems, generating Wikipedia-style articles and working with chemical formulas are included in the repository’s documentation.
GALAI is completely free and open source under the Apache-2.0 license. The model weights are available for research use and there are no paid tiers or hosted versions.
GALAI is for researchers and scientific developers who want a language model specialized for academic and technical content rather than general conversation.