Latest in AI
Anthropic Expands AI for Science to 20 Partners With Free Claude Credits
The research-credit program is moving through foundations and institutions, making the allocation process as important as the model access.
Anthropic announced an expansion of its AI for Science program on October 28, highlighting 20 scientific foundations, institutions and research groups receiving support through free Claude credits. The change extends a program that began with individual researchers into more institutional partnerships, rather than introducing a new scientific model.
For researchers, the important distinction is access. A capable model can still be difficult to use extensively when an experiment requires repeated runs, evaluation and iteration. Credits reduce one part of that cost, although they do not remove the work of building and validating a research workflow.
Which organizations are involved?
The expansion announcement names funders including the Sloan and Simons foundations, Schmidt Sciences and the US National Science Foundation. These organizations will distribute credits to researchers. Institutional partners include the Space Telescope Science Institute, Wellcome Sanger Institute and EMBL.
It also supports specific projects, including Princeton's FlyWire, Columbia's LEAP and the Broad Institute's gnomAD. Anthropic says many collaborations remain early and may produce shared resources such as datasets, software and standards. The announcement does not disclose a total credit budget or a uniform allocation for each partner.
That last omission matters. A list of participating institutions tells us where access is expanding; it does not tell us how much research each allocation can support or which projects will receive it.
How researchers can seek access
Anthropic's existing program guidance directs researchers to an application describing their team and project. It lists scientific merit, potential impact, technical feasibility and biosecurity screening among the evaluation criteria.
The standard application route covers API credits for public models, not access to non-public models or an exemption from usage policies. For a research program with several collaborators, the guidance recommends one application and organization account so credits can be managed across the team.
Researchers should distinguish that existing application route from the new institutional partnerships. Being associated with a named institution does not, by itself, establish that an individual has an allocation. The relevant program administrator is the practical starting point for checking what is available and how to request it.
Why institutional access changes the workflow
A shared allocation creates an opportunity to do more than reimburse isolated experiments. An institution can compare approaches, preserve useful evaluation code and make a successful workflow available to another team. That is an operational possibility, not a result established by this announcement.
It also creates a dependency worth planning for. A workflow that is affordable while credits cover inference needs a cost model for what happens afterward. Recording the calls, models and evaluation steps used in an experiment makes both reproducibility and future budgeting easier to assess.
Free access is an input to research, not evidence of a discovery. The useful outputs to watch are published methods, independently checkable results and resources other researchers can actually reuse.
If these partnerships produce those outputs, their value can extend beyond the teams receiving the credits. That is a more meaningful test than the size of the partner list.
I’m Paras. I work at the intersection of AI, engineering and distribution.
I help technical teams explain complex research and product work through clear writing and launch storytelling.