BBiomni Lab

Biomni Lab is a web workspace where biologists direct AI agents across more than 300 databases, software tools and analysis packages, from literature search to omics analysis and GPU jobs. It is best for academic labs and discovery teams at biotech and pharma companies that want agents to do the computational legwork, and it is unusually cheap to try: the Individual plan is free with no credit card, Pro is $100 a month and academics pay half.

Usage runs on a daily allowance whose size is not yet published, and heavy days can end in a queue, so a published allowance would help users plan. The product is also young, launched as a preview in February 2026, though its open-source predecessor from Stanford is used by more than 50,000 scientists, according to a September 2026 Fireworks AI case study. We rate it 8.7/10 on our scale. This review is based on Phylo’s product, pricing and documentation pages and on published case studies; we have not used it ourselves.

Phylo’s Published Usage and Funding Figures

Phylo is a young company, but it has published more hard figures than most research agent vendors. Each item below was checked on the source page.

For a buyer, the adoption numbers matter because they come from the open-source Biomni agent, which researchers chose to run before any sales team was involved. The pharma names on the enterprise page show that large companies have moved past pilots; Ono’s executive said its researchers quickly adopted the platform.

The Fireworks numbers explain the pricing. By routing most traffic to open-weight models and keeping frontier models for the hardest tasks, Phylo cut its inference costs by 60%, which is how it can offer a free plan with GPU jobs and a $50 academic tier. The 40 times speedup is one user’s report rather than a guarantee. Fast model turnover has two sides: new models arrive within a day of release, and defaults can change, so record the model used for any analysis you may need to reproduce.

An Integrated Biology Environment Run by Agents

Phylo calls Biomni Lab an integrated biology environment. In practice it is a web workspace where you describe a research task and agents carry it out using a large catalog of biological tools, with results returned to a private workspace where you own the IP of what you generate.

From Question to Analysis

1. Pose the question. A scientist describes a task, such as analyzing a CRISPR screen, reviewing literature on a target or exploring public omics data.
2. Agents plan and gather. The agents choose tools from more than 300 databases, software systems and analytical packages, including Consensus for literature, COSMIC for cancer genomics and Addgene for plasmids.
3. Code and compute. They write and run analysis code and launch HPC CPU and GPU jobs, with several tasks running at once on paid plans.
4. Review results. The scientist checks outputs, refines the request and decides what to test in the lab.

Phylo says there are more than 100 prebuilt agentic workflows, and the enterprise page lists use cases built on EHR and claims data, public omics datasets and pharmacovigilance data. Phylo also reports that the Biomni research was published in Science.

Model Choice and Usage

Usage is measured by the model and compute a request consumes, not by message count. A quick question uses far less than a multi-step analysis. Phylo recommends open-weight models for most queries because they consume less usage, and suggests reserving frontier models for tasks open models cannot complete. Phylo adds new models to production within 24 hours of release, according to the Fireworks case study, and sets default models per task type using its internal BiomniBench evaluations.

Enterprise Deployment

Enterprise customers connect internal data, tools and workflows, get unlimited daily usage with expandable compute and storage, and choose multi-tenant, single-tenant or customer-hosted VPC deployment. Deterministic pipelines give reproducible results, which matters for teams that need to rerun an analysis months later. SSO, role-based access and white-glove onboarding come with the plan.

Scientists Still Frame Questions and Run the Wet Lab

The agents do computational work; scientists still frame the question, judge whether an analysis is sound and decide which experiments to run. Biomni Lab does not run wet-lab experiments itself, and Phylo describes experimental services as still underway. Expect to check code, parameters and citations before results feed into a decision.

Individual, Pro, Team and Enterprise Rates

Phylo publishes its prices. Paid plans bill monthly or annually, and the academic discount applies automatically based on your email domain.

A Postdoc’s $600 Year Versus a Lab Team’s Seats

Graduate student or postdoc. The free plan covers occasional analyses. Upgrading to Pro with an academic email costs $50 a month, or $600 a year, for 10 times the daily allowance.

Five-person academic lab. Team with academic seats is $50 per seat, or $250 a month and $3,000 a year, with central billing for the lab.

Ten-scientist biotech team. At $100 per seat, Team costs $1,000 a month, or $12,000 a year. For comparison, a single Kosmos subscription from Edison Scientific is $200 a month, or $2,400 a year.

Variable costs. The daily allowance is not published in numbers, so a team running large analyses every day may hit queues and need Enterprise for unlimited usage. Choosing frontier models over open-weight ones uses the allowance faster. Removing a Team member mid-cycle does not produce a refund, and refunds on monthly plans must be requested within 3 days with minimal use.

Biomni Lab Compared With Kosmos, K Pro by Owkin and Benchling AI

These are the closest alternatives for research teams. Kosmos and K Pro are also agentic research tools; Benchling AI adds agents to a lab data platform.

Biomni Lab is the cheapest way into agentic computational biology for a working lab, and the only one of the four that pairs a free plan with GPU jobs. Kosmos suits long, autonomous runs on a single dataset, K Pro is aimed at enterprise pharma buyers, and Benchling AI makes sense when your data already lives in Benchling. For literature-first work, Elicit is another option. Browse more on our life sciences agents page.

Academic Labs and Computational Biologists

• Academic labs on tight budgets: a free plan and 50% academic pricing make it easy to trial on real projects.
• Computational biologists: agents write code and run HPC jobs across hundreds of tools, saving setup time on each new dataset.
• Biotech discovery teams: Team seats at $100 a month give each scientist Pro-level access with central billing.
• Pharma groups with strict data rules: Enterprise offers VPC deployment, SOC 2, ISO 27001 and no training on customer data.
• Researchers exploring public data: COSMIC, public omics datasets and literature are already wired in; see also our research agents page.

Clinical Ops and Veeva-Based Commercial Teams Have Closer Fits

• Clinical operations teams: trial monitoring and site work are better served by tools such as Medable AI Agents.
• Commercial and regulatory teams in Veeva: Veeva AI Agents work inside those applications.
• Labs that need an electronic notebook or LIMS: Biomni Lab analyzes data rather than managing samples and protocols.
• Teams needing predictable heavy usage on a fixed budget: daily allowances are not yet published, and unlimited usage comes with Enterprise.
• Buyers who want public peer reviews first: there are no ratings on major software review sites yet.

Daily Allowances, Queues and Preview Status

• Daily allowance: requests can be queued or refused after you use the allowance, and the queue itself has limits that Phylo says may change.
• Preview status: Biomni Lab launched as a preview, and team collaboration features are listed as coming soon.
• Model choice affects usage: frontier models use the allowance faster than open-weight ones.
• Verification: agent-written code and analyses can be wrong; review parameters and outputs before acting on them.
• Refund terms: monthly refunds must be requested within 3 days and annual within 14, with minimal use since the charge.
• Vendor metrics: the 40 times speedup and weeks-to-hours results are Phylo and user reports.

Fireworks AI Reports 50,000 Biomni Users

On September 17, 2026, Fireworks AI published a case study on Phylo reporting more than 50,000 Biomni users, 60% lower inference costs and roughly halved time to first token after moving most traffic to open models.

On July 30, 2026, Ono Pharmaceutical announced a collaboration with Phylo to give its drug discovery scientists Biomni Lab agents for reviewing past experiments, analyzing internal data, designing experiments and running computational biology. On February 3, 2026, Phylo launched Biomni Lab with a $13.5 million seed round co-led by Andreessen Horowitz and Menlo Ventures’ Anthology Fund with Anthropic.

Low Prices and a Free Start Make It Easy for Academics

For academic labs and discovery teams doing computational biology, yes. Biomni Lab scores 8.7/10 on our scale. It is free to start, published prices are low for what the agents do, academic users get half off, and the security story includes SOC 2, ISO 27001, no training on customer data and VPC deployment. Pharma adoption by companies such as Bayer, Chugai and Ono, plus more than 50,000 scientists on Biomni, suggest it holds up on real work.

Published daily limits, a release beyond preview and public ratings would lift the score. Start on the free plan with a recent analysis your team has already done by hand, compare results and time, then decide whether Pro or Team seats are worth it. Compare more options on our life sciences agents page.

Biomni Costs and Academic Discounts

How Much Does Biomni Lab Cost?

The Individual plan is free. Pro is $100 a month, Team is $100 per seat a month at a limited-time price, and both drop to $50 for academic users. Enterprise is quoted.

Is Biomni Lab the Same as the Open-Source Biomni Agent?

Biomni started at Stanford in 2024 as an open-source biomedical agent. Biomni Lab is Phylo’s hosted workspace built on that work, adding integrated tools, compute, prebuilt workflows and enterprise controls.

Does Phylo Train Models on My Data?

Phylo says enterprise data is never used to train models and stays within your secure boundary. It lists SOC 2 and ISO 27001, and Enterprise can deploy into your own VPC.

What Happens When I Hit the Usage Limit?

Additional requests may be queued, and if the daily limit or queue capacity is reached, new requests may be refused until usage resets. Pro gives 10 times the Free allowance, and Enterprise offers unlimited daily usage.

Can Biomni Lab Run Lab Experiments?

No. It runs computational work such as database queries, literature review, analysis code and HPC jobs. Phylo describes experimental services as underway, but scientists still run wet-lab work themselves.