Chai Discovery

Chai reports experimental evaluation of 88 designed IgGs across 28 antigens, with 86% showing at most one flagged developability issue and 24 antigens yielding a design with a clean profile; five antibody-antigen complexes were examined by cryo-EM.

Chai's June 2025 release reports laboratory binding validation on roughly 50 antibody targets, almost half yielding validated hits with fewer than 20 candidate designs per target, plus binders on five of five miniprotein targets. This binding-focused research evaluation is distinct from the recorded later IgG developability and cryo-EM study; it is company-reported preclinical evidence, not patient benefit.

Chai applies generative biology models to molecular and antibody design, but broad commercial deployment remains early.

Chai-2 supports de novo antibody and miniprotein design against arbitrary targets and reports double-digit experimental success rates above 10% for antibodies and above 50% for miniproteins, enabling direct characterization in under two weeks without conventional high-throughput screening.

Chai develops advanced biomolecular structure and generative design models, including Chai model research.

Chai develops advanced biomolecular structure and generative design models, including Chai model research.

Chai Discovery's official site states that it has raised US$400 million in new Series C funding. This disclosed completed financing is independent of the historical observation about assembling an AI-biology research organization.

Chai Discovery announced its Novartis collaboration after more than a year of technical engagement, including prior early access to a folding model; the companies extended their work across therapeutic antibody programs. This named commercial collaboration is distinct from financing and the historical general research-team capability.

Chai has assembled a specialized AI-biology research organization capable of model and experimental drug-discovery work.

Chai Discovery publishes the chai-lab Python inference package and Chai-1 code/model weights under Apache 2.0, including configurable MSAs, templates and experimental restraints plus containerized development guidance.

Foundation-model and structural biology skills transfer to adjacent biotech AI roles, with meaningful domain specialization.

Foundation-model and structural biology skills transfer to adjacent biotech AI roles, with meaningful domain specialization.