Introducing Chai-2: Zero-shot antibody design in a 24-well plate

We are opening up early access to academic and industry partners. Sign up for early access here. Check out our technical report for additional information.

Introducing Chai-2: Zero-Shot Antibody Discovery in a 24-well Plate

Today, we're excited to unveil Chai-2, a major breakthrough in molecular design. Chai-2 achieves unprecedented double-digit success rates when designing fully de novo antibodies, increasing hit rates by orders of magnitude compared to prior methods. When testing just 20 designs, the model routinely discovers viable hits across a wide variety of targets. The model’s high success rates and broad generalization paves the way toward a new era of rapid and precise atomic-level molecular engineering.

52 novel antigens targeted by Chai-2. Blue boxes indicate targets with at least one successful binder out of ≤20 assayed designs, representing 50% of the tested targets.

The Challenge of Antibody Design

Antibodies have revolutionized therapeutics, accounting for almost half of all recent biopharmaceutical approvals. However, conventional methods for antibody discovery—like animal immunization or large-scale library screenings—are slow, expensive, and often fail against challenging or novel targets. Previous computational approaches promised efficiency but still required massive experimental screening due to low hit rates.

Enter Chai-2: A New Frontier

With Chai-2, we've developed the first fully zero-shot generative platform achieving double digit experimental success rates in de novo antibody design—a 100-fold improvement over previous computational methods. This leap enables us to bypass the need for traditional high-throughput screening in molecular discovery, and compress hit discovery timelines from months or years down to two weeks. 

Breakthrough Experimental Results

We challenged Chai-2 to design up to 20 antibodies or nanobodies to 52 diverse protein targets, all of which lack existing antibody binders in SAbDab (a subset of the RCSB Protein Data Bank). Remarkably, in a single experimental round we achieved a 16% binding rate, and discovered at least one successful binder for 50% of targets. Beyond antibodies, Chai-2 demonstrates a 68% wet-lab hit rate in miniprotein binder design, routinely yielding picomolar affinities.

A Foundation Model with Atomic Precision

At the core of Chai-2’s success is its multimodal generative architecture, integrating all-atom structure prediction and generative modeling. This enables Chai-2 to create novel and epitope-specific binders across diverse modalities such as scFv antibodies, nanobodies (VHH), and miniproteins. Chai-2 designs are novel and diverse, exhibiting very little similarity to existing antibodies. Chai-2 can also be applied to more complex problems, such as engineering antibodies for cross-reactivity to multiple targets. We further characterized a subset of the hits, finding they are stable, specific, and not polyreactive.

Availability

We are opening up access to a number of academic and industry partners. If you are working on a problem where Chai-2 can accelerate your research, please sign up for early access here.

We will be granting access in accordance with our Responsible Deployment policy, prioritizing access to those working on molecules that positively impact human health and society, while minimizing safety risks.

Looking Ahead - A Shift in Discovery Paradigms

Chai-2’s exceptional success rates mark a paradigm shift in how molecules might be discovered. The shift from stochastic screening to intentional, programmable discovery suggests that antigens once deemed undruggable due to experimental challenges can potentially be addressed by computational design.  On-demand generation of epitope-specific binders could also streamline the development of advanced therapeutic formats such as antibody–drug conjugates, biparatopic constructs, and other multifunctional biologics. Furthermore, by reasoning at the atomic level—including ligands and post-translational modifications—Chai-2’s framework naturally extends beyond conventional biologics to macrocycles, peptides, enzymes, and small molecules.

Chai Discovery's mission is to transform biology from a science into engineering. We believe that our results with Chai-2 mark a key milestone in that journey, pointing toward a transition from empirical discovery to deterministic molecular engineering. Collectively, we see our results as establishing computational-first design as an integral component of modern discovery platforms.

Chai-2 marks a step toward the long-standing aspiration of rational drug design: computationally generating drug candidates that are ready for IND-enabling studies in a single shot, entirely on the computer.

— The Chai Discovery Team