An open-access large language model trained on a massive corpus of scientific papers, formulas, and textbook knowledge.