Anthropic gives free Claude credits to 20 research institutions, while the comments are calling for a new model

Anthropic announced on October 28 that it is expanding its AI for Science program from supporting individual research teams to foundations, research institutions, and research groups. The X account reposted it on the 29th. The core of the program has not changed: giving free Claude credits to high-impact research projects, meaning usage credits for calling Claude.

Post Announcement

The program launched in May 2025. The official description is that it lets researchers use Claude to generate hypotheses, design experiments, analyze complex data, and communicate findings. Over the past few months, it has also offered discounts on the Team plan for scientists and supported hundreds of projects across biology, medicine, physics, chemistry, climate, and mathematics. Here, do not take credits to mean simply free API access. The announcement does not specify whether Claude Code seats are included. One commenter asked directly in a reply: is this Claude Code seats for research groups, or only API credits? The announcement did not answer.

This announcement says it is showcasing 20 partners, while the body names 18. Six foundations and societies: Sloan, Simons, Schmidt Sciences, Open Source for Science Fund, U.S. National Science Foundation, CAS. Six institutions: STScI, Wellcome Sanger, Francis Crick, Imperial College London's Department of Life Sciences, EMBL, Gladstone Institutes. Six projects: FlyWire, LEAP, AstroPath, gnomAD, Tau Ceti, Centre for Population Genomics. The list spans astronomy, genomics, neuroscience, climate, cancer imaging, and mathematical formalization. STScI runs the science operations for the Hubble, Webb, and Roman space telescopes. EMBL-EBI hosts open data resources that biologists use every day, such as Ensembl, PDBe, and Europe PMC. gnomAD aggregates DNA data from more than 800,000 people into an open reference of human genetic variation. Tau Ceti is using Lean to build a community-shared library for autoformalizing mathematics, enabling computers to check proofs.

Two existing collaborations mentioned in the announcement are worth noting. HHMI Janelia and Anthropic are working on Model Hardware Standard to enable AI agents to safely operate laboratory instruments; it is currently a research preview. The Allen Institute is building teams of AI agents to analyze data on the brain, cells, and the immune system; officially, the claim is that this compresses months of manual analysis into hours. One commenter connected the two: on one side, agents operating instruments; on the other, agent teams analyzing data. Once the two loops connect, agents can move from analyzing the previous batch of data to generating the next batch of data. This assessment is more concrete than the vague idea of AI accelerating science.

Another commenter noted that the hardest part for foundations and institutions is domain datasets and evaluation standards, and that the most useful outcome would be shared, reproducible benchmarks for research tasks, not just model access. This point is more crucial than the credits themselves. What research AI often lacks is not a single API call, but comparable tasks, reusable data, and verifiable workflows.

There is also a contrast in the comments. After the announcement, some people were asking for Fable 5.5, others for Opus 5.6, and one commenter wrote, 'blog post day = no new model release day.' In the developer community, a research program still has to compete with model releases for attention. The announcement says many collaborations are still early and that impact will be shared later. What to watch next is how credits are allocated and whether open datasets, open-source software, open standards, and shared libraries can truly last.