GP-MoLFormer represents a breakthrough in molecular generation technology by training an autoregressive transformer on over 1.1 billion chemical molecules.

• Extends chemical language transformers from property prediction to generative tasks
• Demonstrates superior performance in structure-property guided molecular optimization
• Enables efficient discovery of novel molecular structures with desired properties
• Trained on an unprecedented scale of 1.1B+ molecular examples

This advancement accelerates drug discovery by enabling pharmaceutical researchers to rapidly generate and optimize candidate molecules with specific therapeutic properties, potentially reducing development timelines and costs.

GP-MoLFormer: A Foundation Model For Molecular Generation