Claude Science: a research workbench, not a new model
In brief
Anthropic launched Claude Science on June 30, 2026 — a workbench that puts 60+ scientific databases, computing tools, and research workflows in one environment. It runs on existing Claude models, and it's in beta for every paid plan. Here's what it is, who it's actually for, and what it signals about where Anthropic is going.
Contents
Anthropic launched Claude Science on June 30, 2026. It is a workbench for computational research: one environment that connects to 60+ scientific databases, analysis tools, and research workflows, with an AI assistant that coordinates work across them.
The clearest way to describe it: Claude Science is to scientific research what Claude Code is to software engineering. Give it a high-level instruction, and it carries out multi-step work — writing and running code on compute clusters, pulling from databases, and assembling results.
It is not a new model
This is the detail most coverage buried, and it's the one that determines whether you care.
Claude Science runs on existing Claude models — Opus 4.8 among them. There is no special biology model, no gated access tier, no separate capability. What's new is the environment: the connections, the tools, and the workflow layer around the model.
If you were expecting a science-specialised model along the lines of OpenAI's GPT-Rosalind (April 2026, fine-tuned for biological reasoning), this isn't that. It's a product built on models you already have.
What's actually in it
- 60+ scientific databases, connected and queryable from one place — the genetics, chemistry, and protein biology sources used in drug discovery work.
- Compute cluster access, so the assistant can write and run analysis code rather than describing what code you should write.
- An assistant that acts as project manager, coordinating across the databases and tools rather than answering one question at a time.
- Reproducibility as a design priority — results carry their sources so a researcher can verify how a conclusion was reached.
The problem it targets is specific: a computational biologist's day involves moving between a dozen databases, several analysis tools, a notebook, and a cluster terminal. Claude Science collapses that into one surface.
Who it's for
Primarily scientists in molecular and cellular biology and drug development. Anthropic announced it at an event for pharmaceutical executives, biotech founders, and researchers, and the tool coverage reflects that focus.
It can be pointed at other scientific fields, but the database connections are weighted heavily toward life sciences right now.
Availability: beta, included for all paid Claude subscribers — Pro, Max, Team, and Enterprise. Not a separate SKU, not an add-on.
Why this matters even if you don't do science
Three things worth noticing, whatever your job is.
1. This is the "vertical workbench" pattern, and it will spread. Claude Code proved that a domain-specific environment wrapped around a general model beats a general chat interface for serious work in that domain. Claude Science is the second instance of that pattern. If it works, expect the same shape aimed at other domains where practitioners currently juggle a dozen tools — legal research, financial analysis, clinical operations. The lesson for anyone building internal AI tooling: the model is increasingly the commodity, and the connected environment is the product.
2. It's included, not upsold. A flagship product landing inside existing paid plans rather than as a new tier tells you Anthropic is competing on breadth of included capability. That's relevant when you're comparing plan value against Google or OpenAI.
3. Anthropic is using it on itself. Alongside the launch, Anthropic announced it is running its own research programmes into drugs for rare and neglected diseases using the product. Whatever else that is, it's a company betting its own research output on the tool it just shipped.
The competitive picture
Claude Science lands in a three-way race:
- OpenAI released GPT-Rosalind in April 2026 — a model fine-tuned for biological reasoning.
- Google launched Gemini for Science at I/O in May 2026, bundling AlphaFold and AlphaGenome with 30+ databases.
- Anthropic now ships Claude Science with 60+ databases on general-purpose models.
Two different bets. OpenAI is betting on a specialised model. Google and Anthropic are betting the environment matters more than the model. Anthropic's version is the only one included with a standard paid consumer plan.
There is also a business context worth naming: Anthropic is expanding revenue lines ahead of an anticipated IPO. Claude Science widens the set of professional buyers who have a reason to be on a paid plan.
What to do with this
If you work in or near life sciences research, it's in your existing plan — open it and point it at a workflow you currently do by hand across three tools.
If you don't, treat it as a signal about product direction rather than something to adopt. The transferable question is the one it raises for your own organisation: which of our roles spends its day moving between a dozen systems, and what would a single connected environment for that role look like?
That's the question Claude Code answered for engineers and Claude Science is answering for scientists. It's a good question to ask about your own team.
Related reading
- Claude Opus 5 — the current flagship model
- Connectors and skills — how Claude reaches external systems
- What to build with Claude — picking the right first project
- Deep research guide — Claude's research workflow
Sources: MIT Technology Review, Endpoints News, and Anthropic, June 30, 2026.