On October 28, 2026, Anthropic announced that its AI for Science program is expanding beyond individual researchers. The program gives free Claude credits to scientists working on high-impact projects. It now also partners with science funders, research institutions, and major research projects, which hand out credits to their own researchers or use them internally.
Anthropic says the program has already supported hundreds of projects across biology, medicine, physics, chemistry, climate science, and mathematics, and that many of them produce public goods like open datasets, open-source software, and shared libraries.
The announcement lists the new partners, but it doesn't spell out how a researcher actually gets access. This guide does: the four routes in, who's eligible for each, and practical steps to apply.
What changed
The AI for Science program launched in May 2025 as a direct program for individual scientists. Researchers filled in an application, and Anthropic's team, including subject-matter experts, reviewed it.
The expansion adds a second layer: partner organizations that already fund or employ thousands of scientists. Instead of every researcher applying to Anthropic separately, these partners distribute credits through their own channels or use them across their institutions.
The partners fall into three groups.
Science funders and societies (distribute credits to their researchers)
- US National Science Foundation (NSF): the federal agency behind roughly a quarter of federally supported basic research at US colleges and universities.
- Alfred P. Sloan Foundation: funds research and education in science, technology, engineering, math, and economics.
- Simons Foundation: supports mathematics and basic science, including through the Flatiron Institute.
- Schmidt Sciences: focuses on AI and advanced computing, astrophysics, biosciences, climate, and space.
- Open Source for Science Fund (Renaissance Philanthropy): sustains open-source software that science depends on.
- CAS, a division of the American Chemical Society: the organization that indexes published chemistry and maintains the CAS REGISTRY.
Research institutions (use credits internally)
- Space Telescope Science Institute: science operations for Hubble, James Webb, and Roman.
- Wellcome Sanger Institute: genomics, from human genetics to biodiversity.
- The Francis Crick Institute: London's biomedical research institute.
- Imperial College London, Department of Life Sciences.
- EMBL: Europe's life sciences laboratory, including EMBL-EBI, which hosts resources like Ensembl and Europe PMC.
- Gladstone Institutes: a San Francisco biomedical research institute.
Research projects (receive credits directly)
- FlyWire (Princeton): the project that mapped the wiring of an adult fruit fly brain.
- LEAP (Columbia): an NSF center combining machine learning and physics to improve climate models.
- AstroPath (Johns Hopkins): uses methods from astronomy to image tumors cell by cell.
- gnomAD (Broad Institute): an open reference of human genetic variation used in rare disease research.
- Tau Ceti: an open, community-driven mathematics library in the Lean proof assistant.
- Centre for Population Genomics (Garvan Institute and Murdoch Children's Research Institute, Australia): diverse genomic data and rare disease diagnosis.
Who's eligible: the four routes
Route 1: Through a partner funder
Who it's for: researchers funded by, or applying to, one of the partner funders above, such as NSF grantees or Sloan, Simons, or Schmidt Sciences fellows and grantees.
How it works: the funder receives Claude credits from Anthropic and decides how to distribute them to its own research community. Anthropic's announcement doesn't publish each funder's criteria or application process, and each one will run it differently.
What to do:
- Check your funder's website and newsletters for an announcement about Claude or AI credits.
- Ask your program officer directly. They'll know whether credits are available for your grant or program, and how to request them.
- If you develop open-source scientific software, look at the Open Source for Science Fund in particular, since that's its focus.
Route 2: Through a partner institution
Who it's for: staff and researchers at the six partner institutions (STScI, Wellcome Sanger, the Crick, Imperial's Department of Life Sciences, EMBL, and Gladstone).
How it works: the institution uses the credits internally. Access is managed by your institution, not by Anthropic.
What to do: ask your research computing team, IT department, or research office how to get access. If you're at a partner institution and can't find anyone who knows, your department's head of research or data science lead is a good next step.
Route 3: Apply to the AI for Science program directly
Who it's for: researchers at a research institution who aren't covered by a partner, or who need more than their partner offers.
Eligibility: According to Anthropic's original program announcement, applicants should be affiliated with a research institution. Anthropic selects projects based on:
- Your contributions to science (your track record).
- The potential impact of the proposed research.
- How much AI could meaningfully speed up the work.
The original program emphasized biology and life sciences, with examples like analyzing genetic data, understanding complex biological systems, speeding up drug discovery for diseases with large global burdens, and improving agricultural productivity. Anthropic now says the program covers physics, chemistry, climate science, and mathematics as well.
How to apply: use the application form linked from Anthropic's AI for Science program page. Applications are reviewed by Anthropic's team, including experts in the relevant field. Anthropic doesn't publish a fixed credit amount, deadline, or review timeline, so apply when your project is ready rather than waiting for a cycle.
Route 4: The Claude Team plan for scientists
This is separate from the credits program, and it's the fastest way to get your lab using Claude.
Who it's for: principal investigators (or the equivalent) at accredited universities and nonprofit research institutes, working in the natural sciences, mathematics, computer science, engineering, or related fields.
Who it's not for: for-profit companies, contract research organizations (CROs), and industry R&D teams. They need the regular Team or Enterprise plans at standard prices.
What you get, according to Anthropic's Team plan for scientists page:
| Seat | Science price | Regular price | Includes |
|---|---|---|---|
| Standard | $0 a month | $20 a month | Claude Science, Claude Code, Claude Cowork, scientific database connectors and skills, shared projects, single sign-on |
| Premium | $15 a month ($180 a year) | $100 a month | Everything in Standard, plus 5x more usage and higher limits for long analyses |
The fine print:
- Groups can have 1 to 25 seats. Larger groups should contact Anthropic's sales team.
- The discounted price lasts 12 months from sign-up. After that, seats renew at the regular price, with notice to admins first.
- A group with no activity for 90 days may be moved to standard pricing.
- The discount is limited and confirmed only after verification.
- Anthropic says it doesn't train on your conversations, uploaded files, or research data, and admins can set data retention policies.
How to sign up:
- Go to the Team plan for scientists page and select Verify now.
- Confirm your institutional affiliation and briefly describe your research.
- Wait for review. Anthropic says most applications are reviewed within 5 to 7 business days, and you'll get an email when you're approved.
- Once approved, add your team members to Standard or Premium seats.
If your group needs more usage or API credits than the plan includes, you can apply for more through the AI for Science program (Route 3).
Which route should you take?
| Your situation | Best route |
|---|---|
| You hold an NSF, Sloan, Simons, or Schmidt Sciences grant | Ask your program officer (Route 1), and set up the Team plan meanwhile (Route 4) |
| You work at STScI, Sanger, the Crick, Imperial Life Sciences, EMBL, or Gladstone | Ask your research computing team (Route 2) |
| You're a PI at another university or nonprofit institute | Start the Team plan for scientists (Route 4) now, then apply for credits (Route 3) for bigger workloads |
| Your project needs heavy API use, such as running agents over large datasets | Apply directly for credits (Route 3) |
| You work in industry or at a CRO | Not eligible for science pricing. Use the regular Team, Enterprise, or API plans |
How to write a strong credits application
Anthropic doesn't publish a template, but the selection criteria tell you what reviewers care about. Here's how we'd approach it:
- Lead with the science, not the AI. In two or three sentences, explain the research question and why it matters.
- Name the bottleneck. Be specific about what slows the work down today: reading thousands of papers, cleaning messy data, writing analysis code, annotating images, or checking proofs.
- Show exactly where Claude fits. "Use Claude Code to write and test analysis pipelines for our single-cell data" is far stronger than "use AI to speed up research."
- Estimate the gain. Rough numbers help: weeks saved, datasets processed, or experiments made possible.
- Mention public outputs. Anthropic highlights projects that produce open datasets, open-source code, and shared libraries. If yours will, say so.
- Show your track record. Link to your publications, datasets, or software.
- Describe how you'll check the results. Explain how your team will validate AI-assisted work. Reviewers want to see that you understand the limits.
What researchers are using it for
The partner projects give a good picture of where Claude fits in real research:
- Huge reference datasets. gnomAD and EMBL-EBI's resources hold more data than anyone can read. AI helps researchers query, summarize, and cross-reference them.
- Open-source scientific software. The Open Source for Science Fund and projects like Tau Ceti show the focus on maintaining code and libraries that whole fields depend on, which Claude Code is built for.
- Formal math. Tau Ceti builds a mathematics library in Lean, where AI can help draft and check proofs.
- Imaging and analysis pipelines. Projects like AstroPath and FlyWire process enormous image datasets, where AI can help write and debug the code that does the processing.
Anthropic also points to related work on AI in the lab, including a research preview of the Model Hardware Standard, a shared specification developed with HHMI's Janelia Research Campus that lets AI agents safely operate lab instruments.
The bottom line
Anthropic's expansion makes Claude much easier to get for scientists in its partners' networks, and the Team plan for scientists makes it close to free for most academic and nonprofit labs. If you're eligible, start with the Team plan today, ask your funder or institution about partner credits, and apply directly when you have a specific, high-impact project that needs heavy usage.
Want to build these skills first? See our review of building with the Claude API, our guide to Claude Code workflows, and our roundup of the best Claude courses.
Sources: Anthropic: AI for Science expands to foundations and research institutions (October 28, 2026) · Anthropic: AI for Science program (May 5, 2025) · Claude Team plan for scientists
← All AI & Workflows