M³-20M introduces an unprecedented resource containing over 20 million molecules in multiple representation formats, specifically designed to accelerate AI-driven pharmaceutical development.

• Integrates data from existing databases and partially generates molecules using LLMs
• Offers 71 times more molecules than the largest previous dataset
• Provides rich multi-modal representations to support diverse ML approaches
• Creates a foundation for more effective training of drug discovery models

This dataset represents a significant leap forward for the medical community, enabling more robust AI systems for drug design that can potentially reduce development timelines and costs while increasing discovery rates for novel therapeutics.

M³-20M: A Large-Scale Multi-Modal Molecule Dataset for AI-driven Drug Design and Discovery