Chai-2 at a glance

Chai-2 is a series of AI models for computational protein design. Researchers provide a target structure or sequence and design constraints, and the platform generates novel binders such as full-length monoclonal antibodies, VHH nanobodies, scFv antibodies, or miniproteins. It combines all-atom structure prediction with generative modeling so teams can target particular epitopes, formats, antigen states, chemical modifications, and cross-reactivity requirements.

Chai Discovery's initial study tested no more than 20 designs against each of 52 targets and reported a 16% binding rate overall, with at least one binder found for half of the targets. The company later reported full-length antibodies across challenging targets, including GPCRs and peptide-MHC complexes, along with experimental structural and developability measurements. These are company research results, not proof that every target will produce a viable drug.

Chai-2 is an industry and research platform rather than a consumer app. Commercial organizations request access directly, while limited non-commercial academic access is handled through an interest form. Chai does not publish self-serve pricing, and every computational design still needs rigorous laboratory validation and the normal preclinical and clinical development process.