Research access programs

A channel where a provider hands model access to a class of user instead of selling it by the token: free or discounted seats, and grant-style credit awards, aimed at academics and nonprofit researchers. It sits beside the two pricing shapes llm-api-pricing already tracks — the per-token rate card and the speed-provider's flat subscription — and works on a different variable. The rate card asks how much do you use. This asks who are you, and what are you working on.

The two gates

Every program in this wiki so far runs a pair of checks, and they do different jobs.

• The identity gate decides eligibility. Anthropic's scientist program requires a principal investigator or equivalent at an academic or nonprofit institution, verified by form; the PI then adds their lab. The subsidy therefore lands on an institution's research group, not on individuals, which makes it cheap to police and easy to count.
• The domain gate decides what you get. Anthropic tells biology and chemistry researchers they are "limited to our Opus-class models," and its Fable tier blocks professional biology and drug-development queries outright. Access to the top of the ladder runs through a separate, government-partnered program for life-sciences professionals (claude-mythos-5).

The second gate is not a pricing decision at all. It lets a provider hand a subsidized buyer a large discount and a lower capability ceiling in the same sentence.

Why a lab does this

Nothing in the sourced material states a motive, so treat what follows as reading rather than reporting. The obvious ones: adoption among people who will publish about the tool; discovery results the lab can cite (Anthropic points at its own Riemann zeta and protein-design work when explaining the widening); and a class of user whose price sensitivity is high and whose per-seat cost is low. The diffusion-of-innovation framing for that — subsidizing early adopters to seed a population — lives in ../research-wiki.

What this concept is missing

One source, one provider. Whether OpenAI, Google, Mistral or the open-weight labs run comparable programs is unknown here, so nothing on this page should be read as a description of the market. No program has published uptake, renewal terms, or what happens when the free year ends. And no independent account exists of what a subsidized researcher actually gets — the usage limits are stated as multiples ("5×") of an unpublished base.

Related

anthropic-scientist-access-expansion · llm-api-pricing · llm-provider · claude-opus-5 · claude-fable-5 · claude-mythos-5 · anthropic