Living Models
AI infrastructure for understanding living systems, starting with plant genomics and biological sequence data.
Website: https://www.livingmodels.ai/
Cover Block
Public sources
| Field | Value |
|---|---|
| Name | Living Models |
| Tagline | AI infrastructure for understanding living systems, starting with plant genomics and biological sequence data. [Living Models, retrieved 2026] |
| Headquarters | Paris, France [Tech.eu, March 2026] |
| Founded | 2024 [Crunchbase, retrieved 2026] |
| Stage | Seed [Tech.eu, March 2026] |
| Business Model | B2B |
| Industry | Agtech |
| Technology | AI / Machine Learning |
| Geography | Western Europe |
| Growth Profile | Venture Scale |
| Founding Team | Co-Founders (3+): Cyril Véran, Léonard Strouk, Bertrand Gakière [Living Models, retrieved 2026] [Crunchbase Person Profile, retrieved 2026] |
| Funding Label | Seed |
| Total Disclosed | ~$7,000,000 [Tech.eu, March 2026] |
Links
Public sources
- Website: https://www.livingmodels.ai/
Executive Summary
PUBLIC Living Models is building AI infrastructure for biological sequence data, with an initial focus on plant genomics, and it merits investor attention now because it emerged from stealth in March 2026 with a $7 million seed round and a clear commercial wedge into agricultural R&D rather than drug discovery [Tech.eu, March 2026] [Living Models, retrieved 2026] [Diginomica, May 2026]. The company was founded in 2024 in Paris, and its public positioning consistently ties a Paris-Berkeley identity to foundation models trained on DNA, RNA, and other omics data for biological discovery [Tech.eu, March 2026] [Living Models, retrieved 2026].
The core product family is BOTANIC, a set of plant genomic language models built on transformer architectures; public materials describe BOTANIC-0 as a smaller research series and BOTANIC-1 as a larger follow-on family, with the latter presented as trained on DNA from 320 plant species, though the more independently reported early benchmark was training on data from 43 plant species [bioRxiv, February 2026] [Hugging Face, retrieved 2026] [AlphaSignal, retrieved 2026] [Crunchbase News, April 2026]. What stands out is not proven commercial traction but the attempt to build model infrastructure around crop genetics, where seed companies already hold domain data and established breeding workflows, which could make adoption easier than in more regulated biology markets if the models produce usable gains in trait discovery or gene-function prediction [Diginomica, May 2026] [Crunchbase News, April 2026].
The founding team maps cleanly onto that thesis. CEO Cyril Véran is presented as a repeat entrepreneur with UC Berkeley ties, CTO Léonard Strouk brings biochemistry and AI training from École normale supérieure and UC Berkeley, and CSO Bertrand Gakière brings an academic plant-biology background that third-party profiles tie to Paris-Saclay and Université Paris-Sud [SeedQuest, 2026] [LinkedIn, retrieved 2026] [Crunchbase Person Profile, retrieved 2026] [Loop, retrieved 2026].
On capitalization, the only confirmed round is the March 2026 seed, reported at $7 million and associated with backers including Artesian VC, Asterion Ventures, UC Berkeley-linked investors, Kima Ventures, Station F, and others, although source lists vary slightly across outlets [Tech.eu, March 2026] [Crunchbase News, April 2026] [Living Models, retrieved 2026]. The business model is B2B, but as of May 2026 the company was still pre-revenue, had no commercial customers, and was reported to be in discussions with 10 large seed companies as potential development partners, which makes the next 12 to 18 months a test of whether scientific credibility converts into paid pilots, repeat workflows, and evidence that model performance matters inside real breeding programs [Diginomica, May 2026].
Lightly corroborated -- Built primarily from Tech.eu, Diginomica, Crunchbase News, SeedQuest, LinkedIn, and company materials; key company description and funding are corroborated, but several product-detail claims remain company-led or lightly corroborated.
Taxonomy Snapshot
| Axis | Value |
|---|---|
| Stage | Seed |
| Business Model | B2B |
| Industry / Vertical | Agtech |
| Technology Type | AI / Machine Learning |
| Geography | Western Europe |
| Growth Profile | Venture Scale |
| Founding Team | Co-Founders (3+) |
| Funding | Seed, total disclosed approximately $7,000,000 |
How the Company Got Here
PUBLIC
Living Models presents as a very new company, and the public record is still thin enough that chronology matters more than storytelling. The company says it is building "AI infrastructure to understand living systems," starting with plant biology, and its public materials center on genomic foundation models under the BOTANIC name [Living Models]. Crunchbase lists Living Models, formerly ".omics," as headquartered in Paris and founded in 2024 [Crunchbase, retrieved 2026].
The first clear milestone in the public record is formation in 2024, followed by the company operating in relative stealth until its public launch in March 2026 [Crunchbase, retrieved 2026] [Living Models] [Tech.eu, March 2026]. At that point, Living Models announced a $7 million seed round, and its website began foregrounding BOTANIC-1 as a plant genomic model family and linking out to research materials [Tech.eu, March 2026] [Living Models]. The company website identifies three co-founders, Cyril Véran, Léonard Strouk, and Bertrand Gakière, aligning with Crunchbase person profiles for Strouk and Gakière and with Véran's LinkedIn profile, although the company-overview evidence here rests primarily on the company site and Crunchbase [Living Models] [Crunchbase, retrieved 2026].
Lightly corroborated -- Core company identity and headquarters are supported by Crunchbase and the company website, but founding chronology and launch timing are only partially corroborated within the permitted source set.
Product and Technology
MIXED
The product is still best understood as a model family rather than a full commercial software stack. Living Models says it is building "AI infrastructure to understand living systems," starting with plants, and its first product line, BOTANIC, is described as a set of foundation models for plant biology trained on genomic sequence data [Living Models]. Tech.eu similarly reported that the company is building foundation models trained on DNA, RNA, and multi-omics data for biological and agricultural applications [Tech.eu, March 2026].
The public evidence is reasonably consistent on the technical direction, even if the specifications vary by source and vintage. Living Models' website describes BOTANIC-1 as advancing plant genomic representation learning [Living Models], while Hugging Face listings identify both BOTANIC-0 and BOTANIC-1 as plant genomic language model families [Hugging Face]. For the earlier series, bioRxiv reported BOTANIC-0 models spanning 100 million to 1 billion parameters and said the models performed competitively with state-of-the-art baselines in zero-shot and fine-tuned settings [bioRxiv, February 2026]. Separately, Crunchbase News reported that an initial BOTANIC model had been trained on data from 43 plant species [Crunchbase News, April 2026], and later public materials from the company and third-party writeups describe BOTANIC-1 as a larger follow-on series, including claims of 3.2 billion parameters and training across 320 plant species [Living Models] [AlphaSignal] [explainx.ai Blog].
The commercial use case is narrower than the broader "AI for biology" framing suggests, which is probably sensible at this stage. Public descriptions point to seed and breeding companies as the first intended users, with the models meant to support tasks such as predicting gene function, regulatory elements, and trait associations [Diginomica, May 2026] [Living Models]. Diginomica reported that the company was pre-revenue and had no commercial customers as of May 2026, while speaking with 10 large seed companies as potential development partners [Diginomica, May 2026]. That leaves the core technical question less about whether a genomic transformer can benchmark well, and more about whether BOTANIC can convert model performance into repeatable workflows inside crop R&D teams.
Company-stated, unverified -- This section mixes independent reporting from Tech.eu, Crunchbase News, Diginomica, bioRxiv, and Hugging Face with material product specifications that rely on company claims and lightly corroborated third-party summaries.
Where the Demand Sits
PUBLIC
The market matters now because crop R&D is being pushed by climate stress and biological complexity at the same moment foundation-model techniques are starting to show they can work on genomic sequence data, which is the narrow opening Living Models is trying to exploit [Diginomica, May 2026] [TechRadar Pro, March 2026] [Tech.eu, March 2026].
The evidence base here is thinner than investors would usually want for a clean TAM exercise. No named third-party market report in the reviewed sources provides a direct size estimate for AI infrastructure in plant genomics, so the more defensible read is qualitative: Living Models is positioning into the R&D stack used by seed and breeding companies rather than into broad agricultural software or therapeutic drug discovery [Diginomica, May 2026] [Crunchbase News, April 2026]. That matters because the buyer set is narrower, but the workflows are research-intensive and data-rich, which is usually where specialized model infrastructure has a better chance of earning budget.
The demand signals cited publicly are early, but they line up with that wedge. As of May 2026, the company was pre-revenue and had no commercial customers, yet it said it was in discussions with 10 large seed companies as potential development partners [Diginomica, May 2026]. That is not proof of budget conversion, but it does suggest that large incumbents in seed and breeding are at least willing to evaluate genomic foundation models if the tools can improve gene-function prediction, regulatory-element identification, or trait association work [Diginomica, May 2026] [Living Models, retrieved 2026].
The strongest tailwind in the public record is the convergence of three conditions: large biological sequence datasets, improving transformer-based modeling, and a practical need to shorten crop-development cycles under climate volatility [TechRadar Pro, March 2026] [Tech.eu, March 2026]. Living Models and the surrounding coverage consistently frame the opportunity around climate-resilient crops and faster biological discovery, not around generic lab productivity software [Tech.eu, March 2026] [SeedQuest]. That framing is commercially relevant because crop improvement programs often sit between basic research and field deployment, where better prediction can be valuable even before a platform becomes a full system of record.
The adjacent markets are easier to identify than the exact target market size. On one side sits agricultural genomics, including seed-company trait discovery and breeding research; on the other sits the broader market for foundation models in biology spanning DNA, RNA, and multi-omics workflows [Tech.eu, March 2026] [Living Models, retrieved 2026]. A substitute path for customers would be to keep using internal bioinformatics pipelines, academic models, or biology AI platforms aimed at pharmaceuticals rather than crops. The distinction matters because Living Models is explicitly starting in plant science, where sources argue regulatory burdens can be lower and commercialization timelines can be faster than in therapeutics, although that remains a category-level claim rather than a company outcome [Crunchbase News, April 2026].
Regulatory and macro forces cut both ways. Climate pressure and food-system resilience are clear tailwinds for crop science investment in principle, and multiple articles tie Living Models' pitch to climate adaptation in agriculture [Diginomica, May 2026] [TechRadar Pro, March 2026]. But enterprise adoption in this segment is still likely to depend less on frontier-model enthusiasm and more on whether a tool can fit existing seed-company R&D workflows, preserve data confidentiality, and show measurable lift on real breeding questions, which the public sources do not yet demonstrate [Diginomica, May 2026] [CB Insights].
| Market lens | Publicly supported claim | Source |
|---|---|---|
| Direct target market | No named third-party TAM for AI infrastructure in plant genomics surfaced in the reviewed sources | [Tech.eu, March 2026] [Diginomica, May 2026] |
| Initial customer segment | Seed and breeding companies are the stated commercial wedge | [Diginomica, May 2026] [Crunchbase News, April 2026] |
| Early demand indicator | 10 large seed companies were reportedly in discussions as potential development partners as of May 2026 | [Diginomica, May 2026] |
| Adjacent market scope | The company describes its broader platform ambition as foundation models trained on DNA, RNA, and multi-omics data for biology | [Tech.eu, March 2026] [Living Models, retrieved 2026] |
The table underscores the central issue: the buyer category is visible, but the market size is not yet well quantified in public evidence. For now, this reads less like a broad ag software bet and more like a focused platform bet on genomic R&D budgets inside a small set of sophisticated crop-science organizations.
Lightly corroborated -- Section relies on named publisher reporting for customer wedge and demand signals, but no independent third-party market sizing was identified in the reviewed public sources.
Competitive Landscape
MIXED Living Models is positioning itself less as a crop software application and more as base-layer model infrastructure for plant genomics, which places it in a narrower and earlier competitive lane than most agtech software vendors [Living Models, retrieved 2026] [Tech.eu, March 2026].
The closest alternatives visible in public reporting are not a clean list of named direct rivals, but three overlapping groups. First are incumbent seed and breeding companies, the exact buyers Living Models says it is targeting, which already hold substantial internal genomic datasets and established R&D workflows, making in-house model development a credible substitute for buying external infrastructure [Diginomica, May 2026]. Second are broader biology-model efforts, where foundation-model companies trained on DNA, RNA, and multi-omics data could compete for talent, attention, and eventually customer budgets, even if their first commercial wedge is not agriculture [Tech.eu, March 2026] [LinkedIn, retrieved 2026]. Third are open research artifacts, including Living Models' own Hugging Face releases and bioRxiv work, which lower adoption friction for technical users but also reduce the company's ability to rely on model access alone as a moat [Hugging Face, retrieved 2026] [bioRxiv, February 2026].
The defensible edge today appears to be domain specialization rather than distribution. Living Models has anchored its product story in plant genomic representation learning, first with BOTANIC models trained on 43 plant species according to Crunchbase News, and later with BOTANIC-1 described on public sources as trained on 320 plant species [Crunchbase News, April 2026] [AlphaSignal, retrieved 2026] [Living Models, retrieved 2026]. That focus matters because agricultural genomics has different data structures, commercial timelines, and regulatory constraints than therapeutic biotech, and the founding team combines plant biology and machine learning credentials that are unusually aligned with that wedge [SeedQuest, 2026] [LinkedIn, retrieved 2026]. Still, the durability of that edge is not yet proven. As of May 2026 the company was pre-revenue, had no commercial customers, and was still discussing development partnerships with 10 large seed companies, which suggests the present advantage is talent- and positioning-based rather than contractually embedded in customer workflows [Diginomica, May 2026].
The main exposure is that buyers may decide the strategic asset is their proprietary breeding data, not the external model vendor. That risk is sharper here because the reported customer set consists of large seed companies, precisely the organizations with enough data and incentive to build internal capabilities or commission custom work from adjacent AI-biology teams [Diginomica, May 2026]. UC Berkeley SkyDeck, Station F, and the seed syndicate improve credibility and network access, but they do not by themselves solve go-to-market concentration or make Living Models the default channel into crop genomics programs [Living Models, retrieved 2026] [Tech.eu, March 2026]. The company is also exposed to adjacent biology-model efforts with broader scope across DNA, RNA, and multi-omics, because those players may be able to raise more capital and recruit from the same research labor pool before plant-specific infrastructure spending fully matures [Tech.eu, March 2026] [LinkedIn, retrieved 2026].
The most plausible 18-month scenario is a bifurcation between specialist adoption and internalization. Living Models is the likely winner if seed companies decide they need a neutral, technically credible partner to accelerate model experimentation without staffing full internal foundation-model teams, especially if the open-weight approach continues to perform well in enterprise testing environments [CB Insights] [Diginomica, May 2026]. The likely loser if that happens is the status quo of fully internal genomics tooling at seed companies, which could look slower and more fragmented than buying into a dedicated model stack [Diginomica, May 2026]. The reverse scenario is also easy to see. If large seed companies conclude that model performance depends chiefly on their own proprietary datasets and downstream assay infrastructure, then in-house R&D groups become the winner and Living Models becomes the loser, not because the science is weak, but because the value capture shifts to data ownership and integration rather than foundation-model supply [Diginomica, May 2026] [bioRxiv, February 2026].
Lightly corroborated -- Competitive claims rest mainly on company materials, Diginomica, Tech.eu,
Opportunity
PUBLIC
The prize here is unusually large if Living Models can turn plant genomics from a bespoke research workflow into a repeatable software and model layer used by major seed companies, because the company is aiming at a core R&D bottleneck rather than a narrow point tool [Living Models, retrieved 2026] [Diginomica, May 2026].
The headline opportunity is to become the model infrastructure layer for agricultural genomics, first in seeds and breeding, then more broadly across biological sequence analysis. That is a bigger claim than "an AI tool for plant scientists," and the public evidence gives it at least a credible path. Living Models has raised a seed round large enough to fund early model and partnership work, has positioned itself around foundation models trained on DNA, RNA and multi-omics data, and is already in discussions with 10 large seed companies as potential development partners while still pre-revenue [Tech.eu, March 2026] [Diginomica, May 2026] [Living Models, retrieved 2026]. In plain terms, if even a small number of those counterparties convert from exploratory work into embedded R&D usage, Living Models could sit close to the decision loop for trait discovery and breeding prioritization, which is where software can become hard to displace.
The upside path is not singular, and that matters because early platform bets in biology often need more than one route to scale. The company can plausibly grow through direct enterprise adoption by seed companies, through open research distribution that pulls in commercial users later, or through a broader expansion from plant genomics into adjacent omics workflows if its architecture generalizes beyond the initial wedge [Hugging Face, retrieved 2026] [Living Models, retrieved 2026] [bioRxiv, February 2026].
| Scenario | What happens | Catalyst | Why it's plausible |
|---|---|---|---|
| Seed R&D system of record | Living Models becomes the default model layer used by a handful of top seed and breeding companies for gene-function prediction, regulatory element discovery, and trait-association work | One or more of the 10 active seed-company discussions turns into a development partnership, then an internal workflow standard [Diginomica, May 2026] | The company is already targeting seed companies directly, and BOTANIC is explicitly designed for those use cases rather than general biology [Diginomica, May 2026] [Living Models, retrieved 2026] |
| Open model to enterprise wedge | BOTANIC's public model availability drives researcher adoption, benchmarks, and fine-tuning activity that later converts into enterprise deployments and proprietary model services | Continued release and usage of BOTANIC models on public research platforms, plus evidence that open-weight testing works in corporate environments [Hugging Face, retrieved 2026] [CB Insights] | BOTANIC model families are publicly surfaced, and outside sources describe the open-weight approach as part of commercial testing rather than a pure academic release [Hugging Face, retrieved 2026] [CB Insights] |
| Multi-omics platform expansion | The plant-genomics entry point expands into a broader biological foundation-model platform across DNA, RNA and other omics data, supporting additional commercial verticals over time | BOTANIC-1 performance and model scaling establish a reusable architecture that extends beyond the initial 43-species model and plant-only use cases [Crunchbase News, April 2026] [Living Models, retrieved 2026] [AlphaSignal, retrieved 2026] | Company materials and third-party coverage consistently frame the ambition as foundation models for biology, with plants as the first domain rather than the endpoint [Living Models, retrieved 2026] [Tech.eu, March 2026] |
What compounding looks like is a data and workflow flywheel, not a consumer-style network effect. If development partners bring hard biological questions, domain-specific evaluation tasks, and eventually proprietary datasets, the model can improve on the exact tasks that matter for enterprise adoption. Better task performance should make the next seed-company conversation easier, and a wider installed base should in turn create more opportunities to validate model outputs across species and use cases. Public evidence for the flywheel is still early, but two signals are visible: the company is already speaking with 10 large seed companies as potential partners, and BOTANIC has moved from an initial model trained on 43 plant species to a later BOTANIC-1 series described as trained on 320 plant species [Diginomica, May 2026] [Crunchbase News, April 2026] [Living Models, retrieved 2026] [AlphaSignal, retrieved 2026]. That does not prove defensibility, but it does suggest a path where scale in data coverage and enterprise relevance reinforce each other.
The size of the win is hard to mark precisely because no public market-sizing figure or clean public comparable is confirmed in the source set. Even so, the scenario is legible: if Living Models became a standard research infrastructure vendor to global seed companies, or a broader biological foundation-model platform with recurring enterprise contracts, the outcome could be venture-scale in a meaningful way (scenario, not a forecast) [Tech.eu, March 2026] [Diginomica, May 2026]. The key reason is not near-term revenue, which the company did not yet have as of May 2026, but position. A company that sits upstream in discovery workflows, accumulates domain-specific model advantage, and expands from one omics wedge into a wider biology stack can capture more value than a single application vendor, provided the early partnership motion converts into durable deployment [Diginomica, May 2026] [Living Models, retrieved 2026].
Company-stated, unverified -- This section relies on a mix of independent coverage for funding, pre-revenue status, and partner discussions, plus company and platform sources for product scope and model expansion claims [Tech.eu, March 2026] [Diginomica, May 2026] [Living Models, retrieved 2026] [Hugging Face, retrieved 2026].
Sources
Public sources
[Living Models, retrieved 2026] Living Models | AI for Living Systems | https://www.livingmodels.ai/
[Tech.eu, March 2026] Decoding DNA with AI: Living Models emerges from stealth with $7m | https://tech.eu/2026/03/11/decoding-dna-with-ai-living-models-emerges-from-stealth-with-7m/
[Diginomica, May 2026] Being best is good, being useful is better: how Living Models’ BOTANIC AI sets out to tackle climate change | https://diginomica.com/being-best-good-being-useful-better-how-living-models-botanic-ai-sets-out-tackle-climate-change
[bioRxiv, February 2026] BOTANIC-0 plant genomic language models | https://www.biorxiv.org/
[Hugging Face, retrieved 2026] BOTANIC model listings on Hugging Face | https://huggingface.co/
[AlphaSignal, retrieved 2026] BOTANIC-1 is a genomic language model trained on DNA from 320 plant species | https://alphasignal.ai/
[Crunchbase News, April 2026] 5 Interesting Startup Deals You May Have Missed: A Credit … | https://news.crunchbase.com/venture/interesting-startup-deals-mineral-rights-flying-ferry-ai-clean-tech/
[SeedQuest, 2026] Living Models emerges from stealth with $7M seed to build AI foundation models for biology | https://seedquest.com/news.php?type=news&id_article=168064&id_region=8&id_category=2489&id_crop=
[LinkedIn, retrieved 2026] Cyril Véran - Living Models | LinkedIn | https://www.linkedin.com/in/cyril-veran/
[Crunchbase, retrieved 2026] Living Models - Crunchbase Company Profile & Funding | https://www.crunchbase.com/organization/omics
[Crunchbase Person Profile, retrieved 2026] Bertrand Gakière | Crunchbase Person Profile | https://www.crunchbase.com/
[Loop, retrieved 2026] Bertrand Gakière | Loop | https://app.loopcv.pro/
[explainx.ai Blog, 2026] BOTANIC-1 is a later, larger series of models compared to the initial models trained on 43 plant species | https://www.explainx.ai/
[TechRadar Pro, March 2026] Every living thing on Earth runs on the same programming language: How AI foundation models trained on DNA could transform plant biology. | https://www.techradar.com/pro/every-living-thing-on-earth-runs-on-the-same-programming-language-how-ai-foundation-models-trained-on-dna-could-transform-plant-biology
Articles about Living Models
- BOTANIC-1's 3.2 Billion Parameters Aim to Decode Crop DNA — The Paris-Berkeley startup is pre-revenue but in talks with 10 seed companies, betting its open-weight models can predict traits for climate-resilient crops.