GSK signs up for Chai Discovery's AI protein design models after its own wet-lab tests

GSK has signed up to use Chai Discovery's protein-folding and design models, after testing designs from the San Francisco company in its own laboratories. The agreement was disclosed on 9 October 2026 ¹.

What GSK tested

Chai said its models produced designs against GSK's targets "zero-shot without target-specific training", meaning the models were not tuned on data about those particular targets first ¹. GSK then ran the designs through its internal wet-lab testing and found binders to every target it tested ¹.

Christopher Austin, GSK's senior vice president of R&D technologies, said the company "evaluated their models across a panel of diverse targets and saw designs with strong binding affinities" ¹. He presented the deal as an addition to GSK's in-house work: "Their technology complements our own internal models that we continuously retrain against our own wet-lab data, all helping to enable us to accelerate the discovery of potential new medicines." ¹

GSK joins a client list that, this year alone, has grown to include Bristol Myers Squibb, Eli Lilly, Novartis and Pfizer, and Chai has also raised a $400 million series C round ¹. In preprint papers, Chai has linked its second model, Chai-2, to a 16% hit rate in fully de novo antibody design and shown that Chai-2 could design antibodies against six GPCRs ¹. Its newest model, Chai-3, was built to reach high target success rates and binding affinities, including on difficult targets ¹.

What it means for the market

GSK is the fifth large drugmaker this year to strike a pact with Chai, so competing pharma groups are now drawing on the same external design models ¹. For pharma R&D leaders, GSK's approach is a hybrid one: Austin described Chai's models as sitting alongside GSK's own, which it keeps retraining on its proprietary lab data ¹.

The agreement also fits a wider GSK strategy. Chief executive Luke Miels told journalists in May that the "number one priority for the usage of AI is the innovation dimension" ¹. Since then GSK has agreed a $110 million deal with AI-focused biotech Relation centred on "generating large-scale human cellular perturbation data and deploying them into models" ¹.

What remains unknown is the commercial shape of the arrangement. The disclosure gave no financial terms, no count of the targets GSK tested and no therapeutic areas ¹.

Caveats

The evidence so far is preclinical. GSK's binder results come from its own internal testing, and Chai's hit-rate figures come from its preprints ¹. Drug development timelines are long, and the promise of Chai's technology has not yet been borne out by clinical validation for any of its partners ¹.