Latent Labs Announces AI-Generated Antibodies with Low Immune Activation
Latent Labs, which emerged from stealth earlier this year with $50 million in funding, has announced Latent-X2, a generative AI model that produces antibody and peptide drug candidates with preclinical-like developability from the outset, aiming to reduce or eliminate the need for extensive optimization.
According to internal validation, antibodies generated by Latent-X2 bind their targets with picomolar to nanomolar affinities and show low immune activation in ex vivo assays on human blood panels—claimed to be the first such demonstration for AI-generated antibodies.
Latent-X2 was used to generate binders across 18 challenging targets, with success rates in half of them using only 4 to 24 designs per target. In head-to-head comparisons with approved therapeutics, Latent-X2’s designs matched or exceeded developability properties and showed low cytokine release in ten-donor T-cell assays. The model also produced macrocyclic peptides against K-Ras with performance matching hits from trillion-scale screening campaigns, despite using fewer sequences.
Latent Labs presents this as a shift from iterative hit optimization to zero-shot generation of drug-like molecules. The company’s web-accessible platform supports binder generation across VHH, scFv, and macrocyclic peptide formats, and is now available to selected partners upon request via API or browser. Simon Kohl, Latent Labs CEO and former DeepMind researcher, compared Latent-X2’s approach to advances in fields like aerospace and semiconductor engineering, where design now happens computationally before any physical testing
The company also announced that Stefan Oschmann, former CEO of Merck KGaA, has joined its strategic advisory board. Latent Labs is based between London and San Francisco and staffed by ex-DeepMind and AlphaFold contributors.
Latent Labs previously launched Latent-X (now referred to as Latent-X1) in July 2025 as a no-code, web-accessible tool for generating macrocyclic and mini-binder proteins, aimed at academic and commercial users. Both models are part of a broader UK AI-for-science push, with Latent Labs listed among startups supported by the UK’s sovereign AI strategy. The company raised $50 million earlier this year from Radical Ventures, Sofinnova Partners, and AI executives including Dario Amodei and Jeff Dean. NVIDIA has also backed Latent Labs as part of its $2 billion investment push into UK-based AI infrastructure and life science companies.
Cover: Latent Labs
Topic: AI in Bio
