Living Models (formerly .omics)

1. Production rights required without negotiation: BOTANIC weights are provided under a research-only license; production or commercial use requires a separate commercial license.
2. Need for visible, mature enterprise security/compliance evidence: the public site does not list SOC 2 / ISO 27001 / SSO / audit log details.
3. Need for a complete breeding operations suite: if trial management, field phenotyping, and integrated agronomic workflows are required, incumbent ecosystems and specialized breeding tools may be preferable.
4. Lack of model operations capability: foundation models require compute, monitoring, reproducibility practices, and data pipelines; organizations without model-ops capacity may prefer a managed platform.
5. Need for broad public benchmarks and datasets for auditability: some competitors publish extensive benchmarks and datasets; Living Models’ public datasets are not listed on its Hugging Face org.

This section summarizes differentiators stated by the vendor and observable from public artifacts. The language below is neutral and focused on evidence.

• Availability of research-weight model artifacts on Hugging Face, subject to a research-only license and a commercial licensing path for production use.
• Explicit focus on plant genomics and cross-species pretraining as the primary technical approach.
• Public model card and example workflows (tokenizers, embedding generation, example downstream classifiers) that enable technical evaluation by research teams.

Note: the research-only license includes restrictions (for example, prohibitions on commercial use and anti-distillation terms) that limit unrestricted open-source use.

Because there are no accessible product reviews, available market commentary is largely founder and press quotes, the model license and model cards, and ecosystem comparisons. These are not independent customer validations.

Direct quotes available (company/press):

• Tech.eu quotes CEO Cyril Véran: "We train on DNA, RNA, and gene expression to understand the language of life itself," describing the training-data modality focus.
• AgTechNavigator quotes Véran on positioning: "We don’t touch their data; they retain full ownership," describing a customer-controlled data model (this is a company claim, not independently verified).

• AgTechNavigator reports that BOTANIC reportedly matches state-of-the-art across a set of tasks and suggests potential reductions in breeding cycle time; these claims are attributed to press coverage and the vendor and are not independently validated here.
• The Living Models website states that developing a single crop trait takes many years and frames foundation models as a way to reduce parts of that timeline.

Critical evaluation:

• These claims are plausible but not independently validated in the sources accessible for this report. The primary technical preprints or underlying data were not fully accessible within this research environment, limiting deep technical audit.
• Field outcomes depend on genotype-by-environment complexity, trial design, and breeding program realities; strong model performance on benchmarks does not necessarily translate directly to comparable field gains.

Living Models is a techbio company developing foundation models for biology, beginning with plant genomics. Its BOTANIC model family (starting with BOTANIC-0) is described as a genomic transformer trained to learn representations from plant DNA sequences and adapted to downstream tasks such as predicting gene function, regulatory elements, and trait associations relevant to crop improvement and breeding decisions.

In practice, the publicly visible artifacts today are:

• BOTANIC-0 open-weight research models hosted on a public model hub (three sizes: S/M/L) with example code for embeddings and variant scoring.
• A commercial offering implied via licensing (the company indicates commercialization via licensing; the public model license requires a separate commercial license for commercial use).

Primary target: seed and breeding companies, plant genomics R&D groups, and research institutions working on crop traits (yield stability, drought tolerance, disease resistance, etc.).

Positioning (company-described): the offering is presented as foundation-model infrastructure rather than a single narrow genomic selection model. The company highlights pretraining across multiple plant species and intends the approach to support transfer across species and traits, with an emphasis on enabling smaller breeding programs and use cases that may be under-characterized in existing datasets (per press interviews).