Chai Discovery emerges with $30m in funding and a free model to decode the molecular interactions that drive biological processes.
While the AI drug discovery field consolidates in some areas, new players are also entering the market, including Chai Discovery, which recently unveiled its first release: a new multi-modal AI foundation model designed for molecular structure prediction. The six-month-old startup boasts a team with extensive experience from top-tier tech companies like OpenAI, Stripe, Meta and Google AI, and has secured around $30 million in early-stage funding from backers such as OpenAI and Thrive Global.
Chai Discovery’s first model, Chai-1, aims to decode the molecular interactions that drive biological processes and is designed to support various tasks integral to drug discovery. The company says its model facilitates the prediction of proteins, small molecules, DNA, RNA, covalent modifications and other key biological molecules.
Interestingly, Chai-1 can accessed freely through a web interface, even for commercial applications like drug development. Its model weights and inference code are also being made available as a software library for non-commercial applications, demonstrating Chai Discovery’s interest in fostering collaborations with both the research and industrial sectors.
In terms of performance, Chai Discovery claims that Chai-1 has demonstrated promising results, apparently slightly outperforming models including AlphaFold3 and ESM3 in certain benchmarks. The company also revealed that Chai-1 can fold multimers with greater accuracy compared to the MSA-based AlphaFold-Multimer model, making it the first AI model capable of predicting multimer structures at a quality comparable to AlphaFold-Multimer while using only single-sequences.
“Unlike many existing structure prediction tools which require multiple sequence alignments (MSAs), Chai-1 can be run in single sequence mode without MSAs while preserving most of its performance,” said co-founder Joshua Meier, in a thread on X. “In addition to its frontier modeling capabilities directly from sequences, Chai-1 can be prompted with new data, e.g. restraints derived from the lab, which boost performance by double-digit percentage points.”
For instance, antibody-antigen structure prediction accuracy can be doubled by incorporating lab data, potentially transforming the efficiency of antibody engineering. This capacity to integrate new types of data in real time demonstrates the model’s adaptability and potential for broad application across the drug discovery process.
According to Chai Discovery, Chai-1 is just the initial phase of a broader mission to transform the understanding of biology from a purely scientific endeavor into an engineering discipline. As part of this vision, the company plans to develop further AI foundation models to predict and reprogram interactions between biochemical molecules, which form the essential building blocks of life.
