Anthropic Opens Claude to Scientists With 10,000 Free Seats

Anthropic has opened 10,000 Claude Team subscription seats to scientists, most of them free, in the company's largest push yet to put its models directly into working laboratories. The program, announced on 27 August, is aimed at researchers who have so far had to expense their own subscriptions or make do with free consumer tiers.

The terms are unusually generous. Standard seats under the new Claude Team plan for scientists cost nothing. Premium seats, which carry five times the usage limits, run to $15 a month. Anthropic has fixed both prices for a full year, which matters in a field where grant cycles are long and software budgets are fought over line by line.

Eligibility runs through the lab, not the individual. A principal investigator or equivalent at an accredited university or nonprofit research institute verifies through a form, then adds the researchers in their group. The company says the plan covers the natural sciences, mathematics, computer science, engineering and adjacent fields, and that it intends to extend the offer well past the first 10,000 seats in the coming months.

A workbench, not just a chat box

Alongside the seats, Anthropic introduced Claude Science, a product built around how research actually gets done. It folds in the tools scientists reach for most often, produces auditable artifacts so a result can be traced back through its steps, and gives flexible access to computing resources for heavier jobs. The auditable part is the tell. Anthropic is aware that a scientist cannot cite an answer they cannot check, and a black box is worth little at the bench.

Labs that want more than a subscription can apply separately to Anthropic's AI for Science program, which offers up to $50,000 in API credits per project for work the company judges to have real scientific merit.

Why now

The move fits a pattern. Anthropic has spent recent months trying to show Claude earning its keep in serious technical work rather than marketing copy. In August the company reported that Claude had designed working protein binders for 14 of 15 targets, and it has put $5 million into studying how its own tools affect the people using them. Handing scientists cheap, well-instrumented access is the logical next step, and it also seeds goodwill among the exact users who will stress-test a model hardest.

There is a competitive edge to it too. OpenAI and Google have both courted researchers with credits and academic tiers, and getting a generation of graduate students to run their experiments through Claude is a durable form of advantage. For the scientists themselves, the calculation is simpler. A free seat on a frontier model, with a tool that keeps a paper trail, is a good deal on its own terms.