• Latent-Y, the first lab-validated drug design agent, is now open to researchers worldwide through the Latent Labs Platform
• Every approved researcher receives a free daily quota of 250 designs, with additional credits available on demand for larger campaigns
• In published results, Latent-Y reached a 67% target-level success rate across nine targets, with binding affinities in the single-digit nanomolar range
• Researchers at UC Davis, LMU University Hospital and the Translational Genomics Research Institute produced lab-validated designs on the Latent Labs Platform, including groups with no prior computational design experience
• Results include de novo designed functional inhibitors of human ion channels, CAR-T cells that killed tumour cells in vitro and are now in animal studies, and high-affinity binders against a conserved malarial target

LONDON & SAN FRANCISCO, July 15, 2026 — Today, Latent Labs launches Latent-Y, its lab-validated drug design agent, to researchers worldwide through the Latent Labs Platform. Any approved researcher can now access Latent-Y for free, with a daily design quota included at no cost. For larger campaigns, researchers can purchase additional design credits on demand.

Latent-Y is an AI agent for drug design. It takes a therapeutic goal, specified in written language or as a scientific publication, and reasons through the design process end to end, delivering lab-ready sequences. It works through the decisions a specialist structural biologist would make, from identifying the right site on the target to selecting and refining candidates against it. The agent can run without intervention, or pause at each stage for researcher review, giving researchers full control over every design decision. In published results, Latent-Y achieved a 67% target-level success rate across nine targets, with binding affinities reaching the single-digit nanomolar range, compressing weeks of expert workflow into hours, a 56x acceleration.

Ahead of today's launch, the Latent Labs Platform was made available to a cohort of academic research groups, spanning ion channel structural biology, cancer immunotherapy and infectious disease, and including labs with no prior computational design experience. At UC Davis, Prof. Vladimir Yarov-Yarovoy's group designed nanobodies from scratch against a human ion channel. Of five taken into the lab, four blocked the channel's electrical current, and the most potent did so at very low concentrations (an IC50 of 30 nM), a potency usually reached only after rounds of laboratory refinement, but achieved here on the first attempt. At LMU University Hospital, Dr Daria Briukhovetska's group built designed binders into CAR-T cells that efficiently killed tumour cells in vitro and have since moved to animal studies. At the Translational Genomics Research Institute, Prof. John Altin's lab confirmed dozens of high-affinity binders against a conserved site on a malaria parasite protein, one the parasite cannot mutate away without losing its ability to cause severe disease.

Latent-Y is the reasoning and natural-language interface to the Latent Labs Platform, the same models and tools that produced the results above. It runs entirely in the browser, with no infrastructure to set up and no pipelines to maintain, so researchers can start designing immediately and at scale. For scientific teams, this means a single researcher can pursue targets that would otherwise sit out of reach.

To mark the launch, Latent Labs will host a live demonstration and scientific presentation in London, on July 15 2026. Attendees will see Latent-Y run a live design campaign and hear results from the beta program directly from the research teams. The evening will conclude with a panel discussion on “Frontier protein design, now in every researcher’s hands”, featuring the former CEO of Merck KgaA, Stefan Oschmann. Registration is open at www.luma.com/latent-y-launch.

"We built Latent-Y to be a force multiplier for researchers. Just as coding agents are accelerating software development, Latent-Y accelerates drug development: one person running dozens of design campaigns in parallel, each progressing with expert-level biological reasoning. Researchers can now design antibodies, nanobodies and macrocycles at unprecedented scale, all from a natural language prompt," said Simon Kohl, CEO and founder of Latent Labs.

Researchers own the sequences they generate, and Latent Labs does not train on user data or outputs. Access is available at platform.latentlabs.com; applicants provide their affiliation and research use case. Commercial partnerships are available through a separate enterprise tier at partnerships@latentlabs.com.

• Latent-Y is coming to more researchers: Latent Labs will open a Research Tier on the Latent Labs Platform, giving approved researchers access to run large-scale agentic protein design campaigns directly from the browser.
• A live launch in London: on July 15, Latent Labs will host a launch event featuring a deep dive into Latent-Y and first-hand accounts from researchers with early access.
• Designs that work in the lab, from a prompt: Latent-Y is a protein design agent that designs zero-shot antibodies, peptides, and mini-binders, demonstrated, for the first time, to produce successful antibodies in the lab from natural-language prompts.
• Built with leading collaborators: the agent runs on the Latent Labs Platform alongside the Latent-X model family, with accelerated computing support from NVIDIA.

LONDON & SAN FRANCISCO, June 23, 2026 — Latent Labs today announced the upcoming launch of a Research Tier for Latent-Y, its lab-validated autonomous drug design agent, and invited the research community to a launch event in London, on July 15, 2026.

The Research Tier will give researchers accepted into the program broader access to Latent-Y on the Latent Labs Platform, with the ability to run large-scale agentic protein design campaigns directly from the browser, no specialist infrastructure or machine-learning expertise required. Registration for the launch event is open at www.luma.com/latent-y-launch.

About Latent-Y

Latent-Y is an autonomous AI agent for drug design. It is able to design zero-shot antibodies in a range of therapeutically relevant formats, peptides and mini-binder proteins. Given a design goal in plain language it handles the full workflow from prompt to lab-ready molecule: literature and database research, target analysis, epitope identification, molecular generation with Latent-X2, computational validation, delivering lab-ready binder sequences at the end. Latent-Y can run end-to-end autonomously, or pause at each stage for researcher review, keeping scientists in control of every design decision.

In published results, Latent-Y was able to design lab-confirmed antibody binders to 67% of the tested targets, with binding affinities reaching the single-digit nanomolar range, compressing weeks of expert workflow into hours.

The technology behind Latent-Y

Latent-Y runs on the Latent Labs Platform with access to Latent-X2, Latent-X1, external scientific databases, and bioinformatics tools. It can be steered in natural language and works collaboratively or autonomously, completing typical computational protein design workflows in hours rather than weeks. Delivering that speed depends on accelerated computing developed with collaborators including NVIDIA, whose BioNeMo Agent Toolkit libraries, among them cuEquivariance, have helped further accelerate Latent-Y's design pipeline since launch. Looking further ahead, Latent Labs sees fine-tunable scientific models, such as NVIDIA Nemotron open models, as a promising route to agents that learn from wet-lab results.

"We're excited to put powerful agentic systems into the hands of trusted researchers," said Simon Kohl, CEO and founder of Latent Labs. "Just as coding agents are accelerating software development, Latent-Y accelerates drug discovery, giving researchers structure-based protein design directly from a natural language prompt."

Join the launch event

The Latent-Y Research Tier launch event takes place in London, on July 15, 2026. Registration is open at www.luma.com/latent-y-launch. Broader Research Tier access details will be shared at the event and on www.latentlabs.com.

About Latent Labs

Latent Labs builds frontier generative models and autonomous agents for drug design. Its Latent-X family of models and the Latent-Y agent bring structure-based antibody, peptide, and mini-binder design to researchers through the Latent Labs Platform. Latent Labs is based in London and San Francisco. For partnerships, contact partnerships@latentlabs.com.

• Scalable, parallel drug design: one researcher can run multiple campaigns simultaneously across targets and modalities
• From research goal to lab-ready sequences in hours: Latent-Y compresses weeks of expert work into autonomous design campaigns
• Lab-validated results across three antibody design campaigns: achieving a 67% target-level success rate with single-digit nanomolar affinities—without human filtering or intervention
• Powered by Latent-X2: drug-like antibodies with drug-like developability, enabling difficult targets and complex designs

LONDON & SAN FRANCISCO, March 23, 2026 - Today, Latent Labs launches Latent-Y, an AI agent that autonomously designs therapeutic antibodies from a text prompt, compressing weeks of expert work into hours. Powered by Latent-X2, Latent Labs' frontier model for drug-like antibody and peptide design, Latent-Y brings structural drug design to any researcher—no specialist infrastructure required. Latent Labs is opening access to selected partners.

Latent-Y is a force multiplier for drug discovery teams. It operates in the same environment as protein design experts, with access to bioinformatics tools, biological databases, and external publications, applying expert-level reasoning to navigate from research objective to lab-ready candidates. A single researcher can now run multiple design campaigns in parallel, across targets and modalities, each progressing autonomously.

How Latent-Y Works

Give Latent-Y a therapeutic goal specified in natural language, a research work plan, or even a scientific publication, and it proceeds from there. The agent analyses the target molecules, applies biological reasoning to identify viable epitopes, designs antibody candidates using Latent-X2, validates them computationally, and iterates until the design goals are met.

Scientists remain in control throughout. Latent-Y can run fully autonomously end-to-end, or pause at each stage to surface progress summaries and recommended next steps for expert review. Every design decision is recorded reasoning that scientists can evaluate, challenge, and build on.

Lab-Validated Results

Latent-Y has demonstrated biological reasoning with lab-validated success across three distinct antibody design campaigns, without human filtering or intervention:

1. Epitope discovery: Applying biological reasoning to identify epitopes matched to therapeutic specifications, yielding lab-confirmed VHH binders with single-digit nanomolar affinities.
2. Cross-species binder design: Generating antibodies that bind homologous targets across species for translational studies. To execute this campaign, Latent-Y autonomously extended its own capabilities, implementing a custom generative method from a brief natural language description to solve a design challenge it had not been explicitly built for.
3. Design from publication: Processing a peer-reviewed scientific paper to autonomously design antibodies targeting human transferrin receptor (hTFR1) for blood-brain barrier crossing.

In each case, Latent-Y handled the full workflow autonomously, representing the first demonstration of a generalizable protein design agent with lab-validated results across diverse campaign types.

Productivity at Scale

Latent-Y compresses what would take expert teams weeks into hours of autonomous work. In user studies, PhD-level experts working with Latent-Y completed design campaigns 56 times faster than independent expert time estimates. Unlike manual workflows, it can run many campaigns simultaneously. Drug discovery organisations can now pursue more targets, more design strategies, more exploration, with the same resources.

"Latent-X2 gave us the breakthrough: antibodies designed computationally with drug-like developability. Latent-Y builds on that foundation with an expert reasoning layer that handles the full workflow autonomously. The result is speed and scale that weren't possible before—a single researcher running dozens of campaigns in parallel. This is what it looks like when AI becomes a true force multiplier for discovery teams," said Simon Kohl, CEO and founder of Latent Labs.

Access

Latent-Y will be available to selected partners through the Latent Labs platform.

Interest for access can be expressed at partnerships@latentlabs.com.

• Partnership with NVIDIA and Nebius delivers breakthrough performance gains for molecular biology models serving critical research pipelines

GTC — March 2026 — Latent Labs, a leader in generative AI for molecular biology, today announced results from its deployment of NVIDIA Blackwell B300 GPUs in partnership with NVIDIA and Nebius, a leading AI cloud provider for training and inference, achieving over 4x training speedups and more than 60% faster inference for its generative models powering drug discovery.

As Latent Labs' increasingly sophisticated models serve a growing base of industry and academic partners across critical drug discovery pipelines, the demand for compute has scaled in parallel. To meet this moment, the company partnered with NVIDIA and Nebius to deploy its models on best-in-class NVIDIA Blackwell architecture on Nebius’s full-stack AI infrastructure.

The results were immediate and significant. Out of the gate, Latent Labs achieved 2–3x speedups in training workloads. With hardware-specific optimizations tailored to the Blackwell platform, those gains exceeded 4x — a step-change in the pace at which the team can develop and iterate on new models. On the inference side, NVIDIA Blackwell Server Edition GPUs delivered over 60% speedups relative to previous-generation chips, directly shortening the timelines for partners relying on Latent Labs' models in active drug discovery programs.

"These performance gains aren't just benchmarks, they translate directly into faster research cycles and shorter paths to therapeutic candidates for our partners," said Simon Kohl,CEO at Latent Labs. "As demand for our models grows across industry and academia, we're committed to a compute-diverse strategy that meets our partners wherever they are. Blackwell is a significant leap for the workloads we care about most, and our collaboration with NVIDIA and Nebius is an exciting part of that picture." 

The three teams are now looking ahead to deepen the partnership, exploring closer integration between Latent Labs' state-of-the-art generative models and NVIDIA's hardware roadmap, as well as Nebius AI Cloud for model training and inferencing, to unlock further performance gains as both the models and the platform continue to advance. “Latent Labs stands out not only for the strength of its science, but for the culture and operating model behind it. The team moves with remarkable focus and speed, and we’ve been especially impressed by the accuracy and performance of their models in production-grade research workflows. It’s exactly the kind of company we’re excited to support on Nebius AI Cloud,” says Ilya Burkov, Global Head of Healthcare and LifeSciences.

LONDON & SAN FRANCISCO, December 16, 2025 -   Today, Latent Labs announces Latent-X2, a frontier AI model that can design drug-like biologics without iteration. Drug hunters can use Latent Lab's AI platform, built on Latent-X2, to access difficult targets and accelerate development timelines by reducing wet lab work. The generated designs display drug-like properties including low ex vivo immunogenicity, significantly shortening the path from hit to clinical candidate. Alongside the release, Latent Labs welcomes Stefan Oschmann, former CEO of Merck KGaA, to its strategic advisory board. Latent Labs is opening access to the model for selected partners.

The Development Bottleneck. Current wet lab approaches are costly not merely in development effort but in clinical failure. Hits rarely possess properties needed for clinical success, and optimization to address liabilities frequently fails or produces zero-sum tradeoffs. Suboptimal starting points risk costly downstream failure in clinical programs, and addressing shortcomings requires long development timelines.

Latent Labs Platform. Through the Latent Labs Platform, partners and customers can generate antibodies and peptides for disease targets of their choice. It produces high-affinity binders across VHH, scFv, and macrocyclic peptide formats - approaching drug-like quality from the first generation. The platform provides a scientist-friendly workflow, accessible to customers via web browser or by integrating their own systems with the Latent Labs API.

Zero-Shot Antibody and Peptide Design. Latent-X2 generates antibodies that bind challenging targets from the first generation - achieving hits against half of 18 targets selected for diversity and difficulty, with picomolar to nanomolar affinities and each requiring only 4 to 24 designs. The model generalizes beyond antibodies: macrocyclic peptides bind K-Ras, long considered undruggable, matching or exceeding hits from trillion-scale mRNA display screens while testing 11 orders of magnitude fewer sequences.

Drug-Like by Default. Antibodies designed by Latent-X2 exhibit developability profiles matching or exceeding approved therapeutic controls in head-to-head comparison. This extends to proxies for immunogenicity: in the first such assessment of any AI-generated antibody, de novo VHH binders were evaluated across a ten-donor human panel in ex vivo T-cell activation and cytokine release assays, confirming both potent target engagement and low immunogenicity. While animal studies and clinical trials remain ahead, these results demonstrate that AI-generated molecules can now clear preclinical hurdles that previously required lengthy optimization.

"Semiconductors, satellites and aircraft once required repeated build-test cycles, consuming years and billions of dollars. Today they're designed computationally before anything is fabricated. With Latent-X2, drug discovery can move towards that same step change - designing the right molecule from the start," said Simon Kohl, CEO and founder of Latent Labs.

Strategic Advisory Board. Latent Labs announces the appointment of Stefan Oschmann, former CEO of Merck KGaA (until 2021), to its strategic advisory board. "The pharmaceutical industry has spent decades optimizing around the limitations of iterative lab work. Latent Labs is doing something different - building the capability to design molecules that work from first principles. That shift, if it holds, changes the entire logic of drug discovery," said Oschmann.

Access. Latent-X2 will be available to selected partners. Interest for access can be expressed at partnerships@latentlabs.com.
Latent-X2 builds on the success of Latent-X1, released just five months ago. Latent-X1 has been adopted by industry and academic groups worldwide, who value its performance and no-code interface for real lab applications.

Ten months ago Latent Labs announced its $50M funding round, co-led by Radical Ventures and Sofinnova Partners, with participation by Anthropic's CEO Dario Amodei, Eleven Labs' CEO Mati Staniszewski, and Google's Chief Scientist Jeff Dean. The team includes former AlphaFold 2 co-developers and ex-DeepMind team leads, with experience from Microsoft, Apple, Exscientia, Mammoth Bio, Altos Labs, and Zymergen.

• Latent-X generates lab-ready macrocycles and protein mini-binders at all-atom resolution to accelerate drug design
• The model can be accessed through Latent's web-based platform for push-button protein design. Sign ups are open now for early access: platform.latentlabs.com
• Extensive lab validation shows picomolar binding affinities outperforming prior models, with 91-100% hit rates for macrocycles and 10-64% for mini-binders

LONDON & SAN FRANCISCO, July 22, 2025 — Today, Latent Labs is launching Latent-X, a frontier AI model for push button protein design, outperforming competing models under identical laboratory conditions. The model is available for early access on Latent's no-code AI protein design platform, where users can upload protein targets and generate cyclic peptides and mini-binders directly in the browser. Through the platform, users can generate, explore, and score binder designs, selecting top-ranked structures for further lab testing. The platform includes a free tier for both commercial and non-commercial users. Sign up is available at platform.latentlabs.com.

Latent Labs is a frontier AI lab working to transform the expensive, labor intensive, and high failure rate processes of drug discovery into automated drug design. Traditional drug discovery requires screening millions of random molecules—a process where hit rates are typically well below 1% and each experiment takes months and costs thousands of dollars. With Latent-X, drug designers can generate high-confidence binders with the push of a button, achieving what would typically require testing millions of candidates by testing as little as 30 candidates per target.

AI models have recently enabled solutions to previously insurmountable technical challenges in biology. With generative models, frontier AI can go beyond predicting structures to creating new sequences and structures of candidate drugs.

"We envision a future where effective therapeutics can be designed entirely in a computer, much like how space missions or semiconductors are designed today," said Simon Kohl, CEO and founder of Latent Labs. "Our platform empowers scientists with lab-validated protein binder design at their fingertips, whether they're experts or new to AI-powered drug design, and without needing AI infrastructure. This is the first step on our mission toward making biology programmable in order to make drug design instantaneous."

Latent-X generates functional, high affinity de novo binders with breakthrough laboratory performance. In extensive wet lab experiments across 7 therapeutic targets, Latent-X achieved 91-100% hit rates for macrocycles and 10-64% hit rates for mini-binders. The model delivered picomolar binding affinities for mini-binders and single-digit micromolar affinities for macrocycles, with generated binders showing strong target specificity. In head-to-head experimental comparisons, Latent-X exceeded the prior state-of-the-art, outperforming existing generative tools in both in silico evaluations and laboratory validation. Macrocycles are a sought after drug modality for their potential oral deliverability, with their compactness promising tissue permeability while retaining specificity. Mini-binders are a versatile new binder modality that offers high specificity in a flexible format. Full results are available in our technical report: latent-x.latentlabs.com.

The Latent Labs Platform allows users to access the state of the art in protein binder design in an intuitive platform for target upload, hotspot selection, binder design, and computational ranking. The platform features structure visualization, predicted structure overlays, and computational metric rankings allowing to replicate the AI workflows used to generate our successfully lab-validated binders.

Latent-X is a general purpose frontier model that creates binders from scratch for unseen or previously untargeted proteins, solving the geometric puzzle of binding at the all-atom level. The model generates designs over 10x faster than previous methods and co-samples sequence and structure simultaneously, allowing for computational experimentation within seconds. Latent-X generalizes beyond nature's repertoire by generating all-atom binder structures that obey atomic-level biochemical rules, opening doors to other therapeutic modalities that depend on target-specific binding—nanobodies and antibodies being prime examples. The company is now open to partnerships to bring these expanded capabilities to new drug applications.

Only five months ago Latent Labs announced its $50M funding round co-lead by Radical Ventures and Sofinnova Partners, with participation by Google’s Chief Scientist Jeff Dean, Anthropic’s CEO Dario Amodei and Eleven Labs’ CEO Mati Staniszewski. The team consists of former AlphaFold 2 co-developers, ex-DeepMind team leads, and brings rich experience from Microsoft, Apple, Stability AI, Exscientia, Mammoth Bio, Altos Labs and Zymergen.