The workflow combines downloaded model weights and reference data with configurable prediction, sequence-alignment, and template resources. Jobs are described in JSON, and predictions are written as CIF structures with confidence files, allowing developers to inspect and rank results across seeds and samples.
Mol-1 is useful for researchers building reproducible computational biology workflows and adapting structure prediction to their own data. The repository includes examples, input and output documentation, and fine-tuning data preparation, so users can work with the released model rather than only read its technical report.
