Latent Labs
London/San Francisco de novo protein-design model company (Latent-X, Latent-X2, Latent-Y agent) founded by an AlphaFold2 alumnus — sells models, not drugs.
1. Core Product / Service
Latent Labs builds generative models that design therapeutic binders from scratch. Latent-X (2025) designs peptide and mini-binder sequences against a chosen target. Latent-X2 (December 2025) extends to antibodies in VHH and scFv formats plus macrocyclic peptides; the company reports hits against half of 18 hard targets using only 4–24 designs per target, picomolar-to-nanomolar affinities, developability matching approved controls, and low immunogenicity in a ten-donor ex vivo T-cell panel — no animal or clinical data yet [3]. Latent-Y (opened worldwide July 2026) is a design agent built on Latent-X2: it takes a therapeutic goal stated in plain language or as a paper and returns lab-ready sequences, with a free quota of 250 designs per day and paid credits for larger campaigns [5].
Wet lab: small in-house lab plus outsourced scale. At launch (~15 staff) the company described a small San Francisco wet lab beside the computational protein-design team, with models built in London [2] (the SF location rests on one press source). In January 2026 GenScript was announced as the wet-lab validation partner for Latent-X2 — high-throughput CHO and cell-free expression, purification and functional testing — framed by the CEO as scaling capacity beyond the internal lab [4]. There is no owned discovery pipeline and no manufacturing.
Business model: pure model vendor. Revenue comes from (a) self-serve web/API access with free tiers and on-demand credits (Latent-Y), and (b) platform access for selected partners (Latent-X2) [3][5]. No named commercial pharma customer or deal terms had been disclosed as of October 2026; early named Latent-Y users are academic (UC Davis, LMU University Hospital, TGen) [5].
2. Target Users & Pain Points
- Biotech/pharma discovery teams that want de novo binders against difficult targets without running large display libraries.
- Academic labs that lack computational protein-design expertise — the browser interface and agent wrapper are aimed at bench scientists rather than ML engineers [3][5].
Pain solved: replacing billion-scale display screens with tens of designs per target, shortening hit discovery from months to weeks.
3. Competitive Landscape
| Player | Offer | Vs. Latent |
|---|---|---|
| chai-discovery | Chai-2 de novo antibodies; licensing deals with large pharma | Larger and further along on named pharma deals |
| nabla-bio | JAM platform, partnership with upfront + milestones | Deal-based economics vs Latent's self-serve credits |
| evolutionary-scale | ESM3 via API | Team folded into a non-profit in 2025 — a cautionary precedent for pure-model vendors |
| absci / generate-biomedicines | Generative design + own pipeline | Integrated drug developers, not tool vendors |
| adaptyv-bio | Wet-lab binder validation as an API | Complement rather than competitor — the validation end of the loop |
4. Unique Observations
- Free-tier funnel for a biology model is unusual: Latent-Y's 250 free designs per day mirrors developer-tool go-to-market more than biotech BD, betting on adoption before price discovery [5].
- Validation, not design, is the bottleneck: the GenScript tie-up shows that even a model company needs industrial expression and assay capacity to keep claims credible — the lab is outsourced, but it cannot be absent [4].
- Unproven monetization: with no disclosed pharma contract, the model-vendor risk that ended evolutionary-scale's independent path applies here too; partner disclosures are the key signal to watch.
5. Financials / Funding
- Seed + Series A: $50M total announced February 2025, including a $40M Series A co-led by Radical Ventures and Sofinnova Partners; Flying Fish and Isomer joined, with existing investors 8VC, Kindred Capital and Pillar VC [1].
- Angels: Jeff Dean, Aidan Gomez and Mati Staniszewski among others [1].
- No later priced round confirmed as of 2026-10-14 (an aggregator lists ~$59M total, unverified).
6. People & Relationships
- Founder & CEO: Simon Kohl, former co-lead of the protein-design team at google-deepmind and senior research scientist on AlphaFold2 [1].
- Advisory: Stefan Oschmann (former CEO of Merck KGaA) joined the strategic advisory board with the Latent-X2 launch [3].
- Partners: GenScript (wet-lab validation) [4].
- Peers: chai-discovery, nabla-bio, isomorphic-labs, alphafold3.
Sources
- [1] Business Wire, "Latent Labs Secures $50M in Funding…" (2026-10-14)
- [2] TechCrunch, "Founded by DeepMind alumnus, Latent Labs launches with $50M…" (2026-10-14)
- [3] Just AI News, "Latent Labs Launches Latent-X2 AI Drug Design Model" (2026-10-14)
- [4] Newswise, "GenScript … wet lab partner for Latent-X2" (2026-10-14)
- [5] Business Wire, "Latent Labs Launches Latent-Y to Researchers Worldwide…" (2026-10-14)