Recursion is a TechBio company that builds an end-to-end drug discovery platform called the Recursion OS. Its core interface, LOWE (LLM-Orchestrated Workflow Engine), lets scientists run drug discovery workflows using natural language prompts instead of manual lab coordination. LOWE connects proprietary biological data, computational tools, and automated wet labs into a single system.
The platform targets a well documented problem in pharmaceutical research. Traditional drug discovery has roughly a 90 percent failure rate, takes over a decade, and can cost close to 2 billion dollars per approved drug. Recursion uses automated CRISPR-Cas9 screening, high content imaging, and transcriptomics to build large scale Maps of Biology and Chemistry, aiming to shorten the early hit identification stage from years to weeks.
Recursion completed an all stock combination with Exscientia in November 2024, folding Exscientia’s precision chemistry and automated synthesis capabilities into the Recursion OS. Following the merger, the combined company reports more than 60 petabytes of proprietary multimodal data spanning phenomics, transcriptomics, proteomics, ADME, and de-identified patient records sourced in part from partners such as Tempus and Helix.
Computation is handled through BioHive-2, a supercomputer built with NVIDIA that ranked 76th on the 2025 Top500 list. LOWE is used internally by Recursion scientists and strategic partners rather than sold as a self-serve product, which positions it closer to a proprietary R&D operating system than a conventional SaaS tool.
Pricing
Recursion does not publish pricing, free trial terms, or self-serve signup options for the Recursion OS or LOWE. No free plan, usage-based tier, or enterprise price list is available on the company website. Access appears limited to internal Recursion scientists and strategic pharma or TechBio partners, consistent with how the platform is described in company materials rather than marketed as commercial software.
* Disclaimer: Please note that pricing information may not be up to date. For the most accurate and current pricing details, refer to the official website.
Key Features
-
✓
Natural language interface for running multi-step drug discovery workflows
-
✓
Generative chemistry models that design and score candidate molecules
-
✓
Built-in ADMET prediction for absorption, metabolism, and toxicity
-
✓
Automatic scheduling of compounds for physical synthesis and testing
-
✓
MatchMaker tool for identifying drug-target interactions in biological data
-
✓
High-throughput lab automation processing up to 2.2 million samples weekly
Use Cases
Identifying New Therapeutic Targets
Biologists prompt LOWE to explore Recursion’s PhenoMap and surface known targets for a disease area. This helps teams find unexplored biology before committing resources to a therapeutic hypothesis.
Designing Optimized Molecules
Chemists ask LOWE to generate compounds for a chosen target. The generative chemistry models rank candidates while flagging likely toxicity liabilities early.
Assessing Drug Candidate Viability
Researchers request ADMET computations on a compound list. This filters candidates by solubility, safety, and pharmacokinetics before lab work begins.
Automating Wet-Lab Experimentation
Once a compound looks promising, scientists ask LOWE to schedule it for experiments. The request routes automatically to Recursion’s automated labs for synthesis and testing.
Analyzing Drug-Target Interactions
Teams use the MatchMaker tool through LOWE to check how a designed molecule interacts with a target pathway. This validates behavior before advancing a candidate further.
Strengths & Weaknesses
Strengths
Can identify hit candidates in weeks for thousands of dollars, well under typical drug discovery timelines and costs.
Built on one of the largest proprietary biological datasets in the industry, now over 60 petabytes after the Exscientia merger.
Automated lab processes are reported to cut the time and cost of many biological assays by more than 75 percent.
A natural language interface lets scientists without machine learning training run complex computational workflows.
BioHive-2, built with NVIDIA, provides substantial computational capacity, ranking 76th on the 2025 Top500 supercomputer list.
Weaknesses
No transparent pricing, free trial, or public signup process is available anywhere on the website.
The platform is built for internal Recursion scientists and strategic partners, not for public self-serve use.
LOWE depends on Recursion’s proprietary automated wet-lab infrastructure to complete its testing and learning cycles.
No integrations with standard third-party lab information management systems are listed publicly.
Who Is This For?
Drug discovery biologists who map disease pathways and model conditions using CRISPR-based interventions and phenomic data.
Medicinal chemists who design candidate molecules and need fast ADMET and solubility filtering during early discovery.
Computational and machine learning scientists building foundation models on top of Recursion’s biological and chemical datasets.
Pharmaceutical and biotech partners seeking to co-develop programs through Recursion’s platform rather than build in-house infrastructure.
Frequently Asked Questions
What is LOWE?
LOWE is an LLM agent that acts as the natural language interface for the Recursion OS. It lets scientists generate compounds, run predictions, and schedule lab experiments through text commands.
Can outside companies buy access to Recursion OS or LOWE?
There is no public signup or self-serve pricing. Access is limited to internal Recursion scientists and organizations that form a strategic partnership with the company.
How much data does the platform draw on?
Following the completed Exscientia merger in November 2024, Recursion reports over 60 petabytes of proprietary multimodal data across phenomics, transcriptomics, proteomics, and patient records.
What changed after the Exscientia merger?
Exscientia became a wholly owned subsidiary of Recursion, adding precision chemistry design and automated small molecule synthesis capabilities to the existing Recursion OS.
What is BioHive-2?
BioHive-2 is Recursion’s supercomputer, built with NVIDIA. It ranked 76th on the 2025 Top500 list and provides the compute power behind the platform’s machine learning models.
Does Recursion offer any tools to external researchers?
Recursion has released open-source RxRx datasets, and its Phenom-Beta foundation model is available through NVIDIA’s BioNeMo platform for external use, separate from the full LOWE system.
How fast can the automated labs run experiments?
Recursion states its automated wet labs can process up to 2.2 million samples per week, generating large volumes of phenomic imaging data continuously.
Is there a learning curve for using LOWE?
Because commands are entered in natural language, Recursion positions LOWE as accessible to scientists without a machine learning background, though it still requires domain expertise to interpret results.
Does the platform integrate with standard lab management software?
No public integrations with common third-party lab information management systems are listed. LOWE is built to work within Recursion’s own automated lab infrastructure.
What kind of cost or time savings does Recursion claim?
Recursion states that automating biological assays has reduced their time and cost by more than 75 percent compared to traditional testing methods.
Recursion does not list public, self-serve integrations for LOWE or the Recursion OS. Its notable technical partnerships include NVIDIA’s BioNeMo platform, which hosts the Phenom-Beta foundation model for external use, and NVIDIA hardware supporting the BioHive-2 supercomputer. Data partnerships with Tempus and Helix supply de-identified patient records that feed the platform’s biological data layers, though these are data and infrastructure collaborations rather than software integrations available to outside users.