According to Chai Discovery’s co-founders, a biotech firm spent over three years and $5 million trying to generate an antibody for a difficult molecular target. The search involved traditional wet-lab screening and failed to yield results. This year, Chai Discovery claims it solved the same problem in just two weeks using Chai-2, a Gen AI model capable of designing antibodies from scratch with a reported experimental hit rate exceeding 15%.
Founded in 2024, Chai Discovery is developing AI foundation models that simulate and reprogram the interactions between biochemical molecules: an effort aimed at transforming biology into a deterministic design process. The company has now raised $70 million in Series A funding led by Menlo Ventures and its Anthology Fund, a joint initiative with Anthropic. Other backers include Thrive Capital, OpenAI, and DST Global, bringing the total capital raised to $100 million.
Foundation Models for Molecular Design
Chai’s platform builds on the trajectory of generative AI in protein science, following earlier breakthroughs like AlphaFold and EvolutionaryScale’s ESM-3. Its first release, Chai-1, was an open-source model for molecular structure prediction, which founder and CEO Joshua Meier claimed outperformed benchmarks like AlphaFold on tasks critical to drug discovery.
With Chai-2, the company has moved from prediction to generation. The model produces entirely new antibodies based only on a target antigen and epitope, achieving a hit rate the company says is up to 100 times better than previous AI tools. According to Chai, the hit rate for de novo binders is over 15%, compared to 0.1% in older computational methods, and can reach 68% for miniprotein designs.
Chai-2’s architecture integrates atomic-level structure prediction with generative modeling, enabling it to produce diverse formats such as scFv antibodies, nanobodies, and miniproteins. The model can target previously undruggable antigens and design antibodies that are stable, specific, and non-polyreactive, based on in-house experimental validation. The system has already been tested against 52 novel protein targets, and in at least half of the cases, it generated a viable binder from 20 or fewer design attempts.
“We’re seeing a 10% to 20% improvement in success rates on the kinds of problems we apply it to,” Meier said in an earlier interview to Bloomberg. “When we compare to AlphaFold, for instance, we’re finding that our models are consistently better on tasks that are important for drug discovery.”
Strategy, Investors, and Market Position
Chai’s founding team includes Meier, formerly of Absci, Facebook AI, and OpenAI; Jack Dent, an engineering leader from Stripe; and AI researchers Matthew McPartlon and Jacques Boitreaud. Meier was previously Chief AI Officer at Absci, a company that also works on generative models for drug discovery.
“Progress towards game-changing drugs and treatments is far too slow, stymied by costly trial-and-error experiments,” Meier said. “Chai Discovery exists to push the boundaries of what’s possible in this field, applying frontier AI to transform biology from science to engineering, so that breakthroughs can be designed rather than simply discovered.”
Joining the company’s board is Dr. Mikael Dolsten, former Chief Scientific Officer at Pfizer. During his 15-year tenure, Dolsten advanced 150 molecules to clinical trials and oversaw the approval of 36 drugs. “I’m proud to join Chai Discovery and redefine biology from science into engineering,” he said.
Chai’s platform is being made available through early access programs to academic and industry partners. The company says it will prioritize partners focused on human health and responsible use cases, and it plans to use the new capital to extend its platform to harder biological targets and support partner onboarding.
In a competitive landscape that includes DeepMind and EvolutionaryScale, Chai make a technical claim of higher hit rates and practical deployability. Its open-source work with Chai-1, followed by Chai-2’s closed but demonstrated performance, forms a two-part strategy: establish credibility with transparent benchmarks, then build proprietary advantages in rapid antibody generation.
