AI at work

Anthropic Expands Science Credits Through 20 Partners

Anthropic's October 28 science expansion routes Claude credits through research partners. See who benefits, what is known, and how to apply.

Illustrative glass research hub sends light to scientific project symbols beside laboratory equipment.

On October 28, 2026, Anthropic said it is expanding its AI for Science program through 20 scientific foundations, institutions and research groups. The change adds organizations that can distribute Claude credits or use them in their own projects, beyond the program’s original route for individual researchers. Anthropic has not published a credit amount or application schedule for these new partnerships. A researcher should first check whether their institution or funder is involved, then compare the separate individual grant and scientist Team plan routes. The announcement describes the partners and their roles.

What changed on October 28

Anthropic launched AI for Science in May 2025 with free API credits for selected researchers at research institutions. It said applications would be judged on scientific contribution, potential impact and whether AI could accelerate the proposed work. An August 2026 update broadened the fields considered and stated that researchers could apply for up to $50,000 in credits per project through that direct route.

The October announcement adds a distribution channel. Anthropic says it is sharing credits with science funders and societies, which will pass resources to researchers; institutional partners will use credits inside their organizations; and selected initiatives receive credits for specific projects. Its headline figure of 20 describes organizations and groups highlighted across the expansion. The announcement’s three lists name six entries each, while its introduction also discusses Janelia Research Campus and the Allen Institute. It does not say that all 20 were newly added on October 28, or that every partner offers an open application.

Examples show why the distinction matters. The Alfred P. Sloan Foundation, Simons Foundation and US National Science Foundation appear under funders and societies. The Space Telescope Science Institute, Wellcome Sanger Institute and EMBL appear under institutions. FlyWire, the Columbia University LEAP center and the gnomAD database appear under specific research projects. These are different ways to reach researchers, not one public grant with a single form or allowance. Each placement comes from Anthropic’s partner list.

Three routes to research access

The October news is easiest to misread as a new free Claude subscription. Credits and seats solve different problems. Credits can support additional Claude usage for a research project. A Team seat gives a person access to a shared subscription workspace, with usage limits. Anthropic’s scientist offer is separately described on its Team plan page. A group may need both, but one does not automatically award the other.

Route Who it is for Published amount or price What remains to check
Partner-distributed or institutional credits Researchers reached through a named funder, institution or initiative No per-partner or per-researcher allocation published in the October announcement Whether that partner has an application process, eligibility rules or an internal selection process
Direct AI for Science application Researchers proposing qualifying projects Anthropic’s August update says up to $50,000 in credits per project; an award is not guaranteed Current application terms, review outcome and any project-specific allowance
Scientist Team plan Verified principal investigators or equivalent at eligible academic or nonprofit institutions, then their groups Standard $0/month; Premium $15/month per seat for 12 promotional months, subject to verification and availability Final price at sign-up, seat mix, taxes and usage limits

The Team page says the promotion covers 1–25 seats per group, offers 10,000 scientists access at no cost to start, and requires verification. It describes the regular seat prices as $20/month for Standard and $100/month for Premium, but those are comparison prices on Anthropic’s page, not the price of the October partner credits. The promotion is limited and final pricing is confirmed after verification. Treat the current plan page as the source when making a purchase decision.

Which route fits your research group?

If you work at an organization on Anthropic’s October list, ask its research office, program lead or grant administrator whether credits are available for your project and how it allocates them. Being affiliated with a listed institution does not, by itself, establish that your lab has an award. Anthropic did not publish a universal sign-up form for partner credits in the announcement.

If your organization is not listed, the individual AI for Science application route remains relevant. The May launch page invites institution-affiliated researchers to apply, and the August update says any researcher can apply for up to $50,000 per project. Those statements describe a reviewed application, not an automatic credit balance. Check the current form and program rules before planning work against an expected award.

If your immediate need is a shared Claude workspace, look at the scientist Team offer. Its page says a principal investigator or equivalent at an eligible academic or nonprofit institution must verify the group before adding seats. It also says groups on the plan can separately apply for extra usage or API credits. A lab that mainly needs scripted API analysis should not assume that buying Premium seats supplies the same API budget as a research credit award. For an ordinary paid API workload, the AI API cost calculator helps estimate token charges, while the site’s price comparison explains the billed units.

Budget beyond the credits

Anthropic has not stated how many tokens, which models or what expiry date each new partnership will receive. An institution may also set its own allocation process. Until those terms are available, a project budget should separate confirmed awarded credits, projected usage after credits run out, and Team seat costs. The site’s free API tier guide covers public developer offers; the October science program is a selective research route, so its credits should not be entered as a general free tier. The model tracker is useful for model releases, but it does not determine grant eligibility.

For a worked example, five Premium seats at the advertised $15 per seat per month for 12 months would be 5 × $15 × 12 = $900 before tax. That is seat spending if the group qualifies and receives that promotional price. It is not a quote for API usage, an estimate of research credits, or a promise of availability. If a group instead chooses Standard seats under the current promotion, the listed seat price is $0/month, subject to the same verification and availability conditions. Compare this with the site’s Claude Pro versus Max guide only when evaluating personal subscriptions; those plans are another separate purchase.

What remains unknown

The October announcement gives no per-partner credit value, award timetable, number of researchers supported by each partner, or expiry terms. It also does not report comparable measured scientific outcomes for all the listed collaborations. Some partnerships are described as early-stage. Claims that the credits will speed a particular lab’s discovery would go beyond the information published so far.

For now, the useful action is to identify the relevant route and obtain its actual terms before committing a research budget. A named partner can confirm its own distribution process; Anthropic’s direct application can confirm an individual award; Team verification can confirm a seat price. Keep those confirmations separate in the budget.

Sources

All pages read on October 29, 2026.

Free toolWhat will 10,000 chats a month cost?Estimate my bill
DailyTech Editorial Team

We explain AI with numbers you can check. Every price, limit and benchmark links to its source and the date we read it. How we work · Report an error