Xaira must prove clinic-ready models

BigHat shows that the hard part is not only inventing new antibodies on a model, it is turning them into drug candidates that survive the messy realities of manufacturing, stability, payload attachment, and human dosing. Its closed loop is built around that last mile. The platform makes and measures thousands of variants each week, then feeds those lab results back into the next design cycle, which is how an optimization engine can push a molecule all the way into Phase I.

• BigHat reached a concrete clinical milestone when it dosed the first patient in September 2026 with BHB810, a CDH17 directed ADC for gastric and other GI tumors. That matters because ADCs add another layer of real world optimization, the antibody has to bind the target well and also carry a toxic payload safely.
  1 bighatbio
• The company pairs that clinical progress with pharma validation. BigHat lists completed collaborations with Merck and an active Lilly collaboration, which suggests large drugmakers value a system that improves developability and function, not just one that generates novel sequences.
  2 bighatbio 3 bighatbio
• That is the practical contrast with model first peers. Absci emphasizes de novo generation speed and wet lab scale, while Generate has won credibility by advancing a broader clinical pipeline, including a Phase 3 anti TSLP antibody. The market rewards whichever stack can repeatedly turn design work into human data.
  4 absci 5 generatebiomedicines 6 generatebiomedicines 7 amgen

Going forward, more AI drug platforms will converge on this shape, model plus automated lab plus translational execution. The winners will be the groups that can show a steady handoff from design to IND to patient dosing. For Xaira, that raises the bar from having frontier models to proving those models can produce clinic ready molecules on a repeatable basis.