Chai Discovery, a biotech company focused on molecular AI, has unveiled a major leap in drug discovery. Its new model, Chai-2, can design therapeutic antibodies entirely from scratch, achieving success rates previously thought impossible. The company is based in the U.S. and backed by major investors, including OpenAI and Thrive Capital. In trials, Chai-2 delivered validated antibodies for nearly half of 50 biological targets, often needing just 20 test designs each.
That’s a stark contrast to traditional methods, which often yield success rates below 0.1%. Pfizer’s former Chief Scientific Officer Mikael Dolsten called the results “incredible,” citing potential to speed up drug development timelines. The model is already making waves in the biopharma industry, where slow R&D and high costs often delay critical treatments. Chai Discovery plans to open early access to Chai-2 for selected partners in the coming months.
AI System Designs Antibodies From Scratch in Just Days
Chai-2 is the company’s latest AI model, built to design antibodies de novo, meaning from scratch, with minimal input. Researchers only need to feed the model a biological target and its epitope, the specific site to which an antibody binds. In return, Chai-2 produces full antibody blueprints, including the most complex parts known as CDRs.
These blueprints can be tested directly in labs, without months of trial-and-error. In many cases, antibodies created by Chai-2 showed strong binding, good drug-like properties, and fast development timelines. Some were ready for lab validation in under two weeks. Joshua Meier, co-founder of Chai, described the system as “Photoshop for proteins.”
Meier added, “It enables precise, rapid, and intuitive creation of biologic therapeutics, which can enable previously intractable targets.” This shift challenges older methods like immunization or yeast display, which are often slow, costly, and hit-or-miss. It also improves on previous AI efforts, which had poor hit rates and still needed heavy lab refinement.
Chai-2 Lab Results, Speed Gains, and Industry Impact
The company tested Chai-2 across 50 antibody targets. Nearly 50% of these produced strong, validated hits with fewer than 20 candidate molecules per case. That’s hundreds of times better than the current industry average, which often falls short even with thousands of designs tested. In one case, Chai-2 solved an antibody problem that had previously cost $5 million in research. The AI delivered a working design in hours, which was confirmed in the lab in two weeks. It also succeeded with 5 out of 5 miniprotein targets, a category where other systems have struggled.
This suggests Chai-2 can unlock new areas of drug research. However, the broader use of AI in molecular design still faces trust issues. Critics worry about over-reliance on models and the lack of transparency in training data. Still, Chai’s results are drawing praise from veterans in pharma, including Dolsten, who emphasized the speed and quality of outputs. The company plans to expand access gradually, working closely with early partners to refine use cases.
AI’s Role in Drug Discovery Enters a New Phase
With Chai-2, AI has moved from an assistive tool to a full-fledged designer in the world of drug development. That shift marks a turning point for how new medicines could be created, faster, cheaper, and with fewer dead ends. As biopharma firms face mounting pressure to cut costs and deliver new therapies, tools like Chai-2 offer a compelling alternative. The potential to skip months of lab work could reshape how treatments reach patients. For Chai Discovery, the next step is collaboration. By opening early access to Chai-2, the company is betting on speed, scale, and scientific trust to drive the future of drug design.
About The Author
