Anthropic: Expanding our support for scientists

Anthropic's announcement of 27 August 2026: 10,000 subsidized Claude seats for researchers, a widened credits program, and — in the last paragraph — a statement of which models scientists are allowed to buy at all. First-party and promotional, so T1 for the facts about Anthropic's own programs and worth nothing as evidence that the programs work.

For this spoke the post is unusual. Every other pricing page here reads a rate card. This one prices access to a class of buyer, and then gates that access by the buyer's field.

What is on offer

Claude team plan for scientists — 10,000 seats, one year:

Anthropic says it intends to "extend this program well beyond the initial 10,000 seats." The gate is an identity check: you must be a principal investigator or equivalent at an academic or nonprofit research institution, you verify through a form, and once verified you add your own lab members. So the unit of subsidy is a lab, not a person.

AI for Science credits — previously aimed at biology, now open to "other scientific fields as well, including those working on ambitious, compute heavy-research." Up to $50,000 in credits per project, and here the eligibility is wide: "any researcher is eligible to apply." The path Anthropic describes is a ladder — burn through the subscription, then apply for credits.

Claude Science — the product the programs feed, launched June 2026, described only as integrating "the tools that researchers most commonly use," producing "auditable artifacts," and providing "flexible access to computing resources." No page in this wiki covers it; this sentence was all the evidence we had until claude-nine-loop-amplitude-von-hippel (25 Sep 2026), which shows the harness running Fable 5.1 largely unsupervised for hours-to-days on a physics calculation — and shows the credit mechanism described here in use: Lance Dixon, the outside physicist who validated that result, was paid in Claude usage credits for doing so.

The work cited as the reason for widening beyond biology is Anthropic's own: progress on the Riemann zeta function, and Claude's work on protein design (the protein result is the Mythos-only headline recorded on claude-mythos-5).

The part that reorders this spoke's model of the line

Three sentences at the end of the post do the work:

• Researchers in biology and chemistry "will still be limited to our Opus-class models."
• Fable models "continue to block professional biology and drug development queries" because of dual-use risk.
• Anthropic is "working in partnership with the US government to establish an access program for life sciences professionals to use Mythos-class models," and has "now enrolled our first participants."

Read against claude-mythos-5, which recorded a biology trusted-access program as planned in June, this is that program moving from plan to live enrollment — the first evidence in the corpus that the ungated tier is actually reaching anyone outside the cyber partners.

Read against this spoke's price ladder it cuts deeper. The ladder here sorts Sonnet 5 → Opus 5 → Fable 5 by price and capability. A biologist cannot move along it. For that buyer Opus-class is not the mid rung, it is the ceiling, and the two tiers above are reachable only by a government-mediated program with an unstated queue. The people most likely to want frontier capability in biology are precisely the ones who cannot purchase it.

What the post does not say

• What "5× usage limits" is 5× of. No token, message or hour figure for either seat.
• What happens in year two. The subscription is stated as one year; renewal terms are absent.
• The size of the commitment. No total dollar value, no count of AI for Science grants awarded to date, no acceptance rate.
• Whether the two programs interact beyond the described ladder — whether credits can be spent on the models a biologist is barred from, for instance, is unaddressed.
• How "professional biology" is judged. The Fable block is a classifier decision on a query; the post does not say how it separates a professional query from a student's.

All figures are first-party and dated 2026-08-27. Program terms are volatile — date any quote.

Related

research-access-programs · llm-api-pricing · claude-fable-5 · claude-opus-5 · claude-mythos-5 · anthropic · claude-nine-loop-amplitude-von-hippel · lance-dixon · synthesis