Prima Mente is building biology foundation models on proprietary multimodal multi-omics data generated in-house from its tissue bank, and its Pleiades model family (90M / 600M / 7B parameters, trained on 1.9 trillion tokens at single-nucleotide resolution) hit 0.89 accuracy for early-stage Alzheimer's detection — rising to 0.97 combined with protein tests. The wedge is closing the loop between wet-lab data generation and foundation-model training in a single vertically-integrated company, rather than competing with third-party model shops that scrape public omics.
The team compounded: DCVC, Antler, and B Capital in earlier rounds; UK Sovereign AI Fund and Gates Ventures endorsed the Alzheimer's work; NES committed $10M as compute partner into the Series A; existing investors put in another $10M+. Devin Gilliam (co-founder) and Ravi Solanki (technical, Cambridge / Frontier VC background) run the SF + London operation. SH's role is to add to the $50M target Series A now that $20M is committed — SH is being brought into the remaining $10–20M of allocation.
The Alzheimer's diagnostic and therapeutic market is $10B+ growing double-digits, sitting inside a $67B neurodegenerative market by 2031 (Web). Prima Mente's positioning as "foundation models for biology" places it alongside AlphaFold, RFdiffusion, and Recursion — the AI-x-bio category is currently pricing at $500M–5B for Series A/B biotech-plus-model companies, so a $150–180M pre on real SOTA benchmarks is priced within the category.
| Player | Positioning | Funding / Stage | Edge vs. them |
|---|---|---|---|
| Prima Mente | Vertically-integrated biology foundation models + tissue-bank data generation | Series A, $150–180M pre | — |
| Recursion | Public AI-x-bio drug discovery | NASDAQ:RXRX | Recursion is broader / cellular imaging; Prima is epigenome + neuro-focused |
| Genesis Therapeutics | AI molecule design | Series B, $200M raised | Genesis does chemistry; Prima does biology |
| Evozyne / Cradle / EvolutionaryScale | Protein / biology foundation models | Series A/B | Different modality — protein language vs. Prima's epigenome/single-nucleotide |
| Insitro | ML-driven target discovery | Late-stage, $600M+ | Insitro is target-focused; Prima owns data-generation flywheel |
Moat: The vertical integration is the moat — most AI-x-bio startups rent data; Prima generates proprietary multimodal data through its tissue bank, feeding a per-token training advantage that compounds. Combined with the Doudna-tier team + NES compute partnership, this is a defensible position.
Comparable exits: Recursion IPO ($2B debut, 2021), Insitro / Alphabet Verily and Google Deepmind's Isomorphic Labs spin-out valuation — AI-x-bio pricing continues to expand. Big-pharma acquirers (Novartis, Roche, Lilly) are actively acquiring foundation-model biotechs — Merck's acquisition of Prometheus for $10.8B (2023) is the strategic exit comp for a company with a strong single-indication proof point.
Follow-up conversation as Prima Mente approaches Series A close; SH allocation is under active discussion.
First deep dive with Prima Mente after Pleiades public launch. Team walked through the epigenome foundation model architecture and the Alzheimer's benchmark.