Living Models Secures $7 Million in Seed Funding to Revolutionize Plant Genomics with AI
Living Models, a Franco-American company specializing in AI for plant genomics, has officially exited its stealth mode, announcing a successful $7 million seed funding round. This capital will accelerate the development of transformative AI foundation models aimed at significantly shortening crop breeding timelines and democratizing access to genomic selection, extending opportunities beyond the largest seed corporations.
The funding was co-led by Asterion Ventures and The Galion Project, alongside contributions from Kima Ventures, STATION F, and various strategic investors. In conjunction with the announcement, Living Models has also released its inaugural technical report and made its BOTANIC foundation model available on Hugging Face for the scientific community to explore.
Introducing BOTANIC: Advanced AI for Plant Genomics
Living Models’ flagship technology, BOTANIC, consists of transformer models trained on genomic sequences spanning 43 plant species. With up to 1 billion parameters, BOTANIC has demonstrated performance on par with state-of-the-art genomic models across 22 benchmark tasks—achieved while being trained on only eight NVIDIA H100 GPUs.
The newly acquired funding will allow for a significant upgrade in computational capacity, transitioning to a dedicated NVIDIA B200 cluster with 120 GPUs. This expansion is expected to facilitate the development of larger models, enhance predictive accuracy, and broaden research into genomic domains beyond plants.
Revolutionizing Crop Breeding Cycles
CEO and co-founder Cyril Véran emphasized the pressing need for acceleration in crop breeding cycles, which typically span 8-12 years, a timeline that is increasingly inadequate in the face of climate change. He stated, “A breeder’s workflow is largely defined by what they can afford to phenotype. While genomic selection has brought improvements, the existing models remain narrow, crop-specific, and heavily reliant on data.”
A New Paradigm: From Correlation to Understanding in Genomics
According to Véran, BOTANIC’s unique capability lies in its ability to translate biological structures across species and traits. “By pre-training on 1,600 genomes across the plant kingdom, it encodes profound biological structures that can transfer across different species,” he explained. This allows breeders to execute higher-confidence early-stage selections and reduces reliance on extensive field cycles.
More crucially, Véran identified that BOTANIC transcends conventional statistical pattern-matching, moving towards a more comprehensive understanding of genomic traits. “Standard genomic selection treats markers as interchangeable statistical surrogates. BOTANIC, on the other hand, learns functional genomic signatures tied to specific traits, providing breeders with more than just numbers—it offers biologically informed insights.”
Cutting Down Breeding Time: Potential Gains
When asked about the potential for commercial acceleration, Véran noted existing academic literature. He stated improved early-stage genomic predictions could eliminate one to three field cycles, potentially shortening breeding timelines by 1-4 years, thus helping climate-adapted varieties reach the market more swiftly while ensuring they have well-characterized performance across diverse environments.
Véran believes unique genetic traits like yield stability, drought resistance, and disease resistance are poised to see benefits early on. BOTANIC’s embeddings reportedly act as guides in identifying genomic regions that boost performance in varying environments, transforming traditional black-box predictions into biologically actionable intelligence.
Widening Accessibility to Breeders Worldwide
A critical aspect of Véran’s vision is making these advanced genomic tools accessible beyond the top five seed companies. “BOTANIC provides robust predictions even where internal datasets are sparse. It caters to both a global seed company’s elite germplasm program and a regional breeder focused on locally adapted varieties,” he highlighted.
In North America and Europe, initial commercial traction is gaining momentum, bolstered by academic collaborations with organizations such as INRAE and the University of Florida. Discussions are currently in progress with one of the leading five seed companies, as well as several regional breeders. However, the long-term aspirations remain focused on underserved crops like sorghum, cassava, and millet, ensuring that even smallholder breeding programs in sub-Saharan Africa can reap benefits akin to those in commercial agricultural hubs.
Maintaining Data Privacy through a Licensing Model
Living Models plans to commercialize BOTANIC through a licensing model that allows seed companies and research institutions to personalize AI applications within secured environments. “We don’t access their data, ensuring they maintain complete ownership,” Véran reassured. This approach is designed to counter the challenges faced by existing modeling tools that struggle to derive value from existing data.
By co-developing integrations directly with clients, Living Models aims to ensure BOTANIC seamlessly fits into existing R&D workflows. “Biology presents an information challenge at every level, from individual cells to entire ecosystems,” remarked Léonard Strouk, CTO and co-founder. “While genetic data abounds, the missing piece was an architecture capable of scaling learning from that data.”
“All life on Earth speaks the same genetic language: DNA translates to RNA, which in turn influences proteins, ultimately determining phenotypes,” added Bertrand Gakière, VP of Biology. “We’re not just building another chatbot; we’re creating a model that interprets this genetic code, offering invaluable insights into biological systems.”
