In an era where billion-dollar startups flex trillion-parameter models, a nimble two-person team at OpenProtein.AI has quietly upended the biotech playbook. PoET-2, a breakthrough protein language model, emerges as a David among Goliaths—delivering staggering performance with just 182 million parameters, a feat once thought possible only with astronomical compute budgets and multi–hundred-million-dollar investments.

Built on a cutting-edge multimodal architecture, PoET-2 simultaneously digests protein sequences and structures, reducing the experimental data needed for protein engineering by an impressive 30-fold. This radical approach is set to turbocharge the development of therapeutic antibodies, industrial enzymes, and other custom-engineered proteins, marking a seismic shift in computational protein design.

"PoET-2 represents a fundamental shift in how artificial intelligence learns from protein sequences," said Tristan Bepler, CEO and co-founder of OpenProtein.AI. "Instead of relying on massive compute resources to memorize sequences, PoET-2 learns the underlying principles of protein evolution and function, enabling accurate predictions from minimal experimental data."

Key capabilities include:

● Superior performance with 500x fewer parameters than competitors

● 30x reduction in experimental data needed for protein optimization

● First model to effectively predict complex mutations and insertions/deletions

● Interfaces with proprietary protein databases without retraining

● De novo protein design with programmable constraints
‍

Demonstrating its prowess on industry-standard benchmarks like ProteinGym and CASP15, PoET-2 outclasses traditional methods while demanding only a fraction of the computational resources. This breakthrough is not just a technical win—it signals a broader shift in how AI can democratize complex fields like biotechnology.

"The implications for the democratization of biotechnology are immense," said Tim Lu, MIT Professor, Senti Bio founder, and biotech industry veteran. "Rather than keeping AI technologies in closed ecosystems only accessible to the few, PoET-2 enables highly accessible, affordable, and efficient protein engineering for a wide range of applications, such as antibody optimization, enzyme engineering, and others."

Now available immediately through the OpenProtein.AI platform, PoET-2 offers free access to academic researchers. Commercial users can unlock the full suite of protein engineering tools—including zero-shot variant prediction, ML-guided optimization, and high-throughput APIs. For an in-depth look at this groundbreaking model, read the PoET-2 white paper here: https://www.openprotein.ai/a-multimodal-foundation-model-for-controllable-protein-generation-and-representation-learning