Anthropic on Tuesday announced Claude Science, an AI workbench that gives scientists a single environment to conduct computational research and eliminates the hassle of navigating between databases, pipelines, and tools.
To be clear, Anthropic says Claude Science is “not a new AI model, nor is it a more capable model for biology. It runs the same Claude models that are already available to everyone today (including Claude Opus 4.8) without any special access or gates.”
This workbench builds on Claude for Life Sciences, which Anthropic launched in October 2025. This is essentially an enhancement to the Claude chatbot by improving life science tasks. Claude Science is a dedicated place to do that work.
The launch, announced Tuesday at an AI for Science briefing, fits into Anthropic’s broader push to move beyond model providers and own more of the operating layer for specific industries, much like how Claude Code has become the operating layer for software development. Anthropic is increasingly betting its growth on vertical, workflow-level products rather than just raw model capabilities (which can determine how it competes with its competitors and at what price).
Here’s how it works: One of the main AI assistants acts like a project manager for scientists. Connects to over 60 scientific databases and comes with pre-built toolkits for specific areas such as genomics, protein structure, and chemistry. That assistant can create sub-assistants to help project leaders delegate tasks to experts, or users can delegate work to custom “expert” assistants they build for their own research. A separate fact-checker AI then double-checks citations and calculations before publication.
The fact-checking step is important because as AI-based writing increases, fabricated citations and unverifiable statistics will find their way into papers. That said, it is still the same underlying model check itself and is not an independent source of truth.
Claude Science has other ways to ensure reproducibility, Anthropic said. For example, Workbench can generate diagrams such as 3D structures of proteins and chemical drawers, along with the code that created them. The company says each diagram includes “the exact code and environment in which it was generated, a plain explanation of how it was created, and a complete message history.” This process also saves time by allowing scientists to edit diagrams in plain language and prompting agents to edit their own underlying code.
Another way Claude Science can save scientists time is by running data in the lab’s own infrastructure settings, rather than sending it to Anthropic’s servers.
Early users say they are already putting this to work. Jérôme Lecoq, a neuroscientist at the Allen Institute, used this tool to build a multi-agent computational review pipeline. Stephen Francis’ group at the UCSF Brain Tumor Center leverages Claude Science to speed up comprehensive germline analysis of gliomas by a fraction of the time previously possible, and the results have been independently verified.
Claude Science’s announcement comes months after OpenAI approached the same problem from a different angle. In April, OpenAI released GPT-Rosalind, a specialized model fine-tuned for biological inference.
The differences between the two approaches not only affect whether a specialized model is required, but also who can access it and at what speed. Rosalind was launched as a research preview limited to qualified commercial customers in the United States and is gated through qualification and safety review. Partners including Amgen, Allen Institute, Moderna, Thermo Fisher, and Novo Nordisk gained early access.
And then there’s Google DeepMind, which is playing a completely different game. DeepMind actually owns basic scientific models like AlphaFold and AlphaGenome, but the other two can only be called tools. Its Gemini for Science platform bundles these and more than 30 life science databases into one skill set.
The result is that three very different distribution strategies are competing for the same scientific research market. Anthropic has expanded broadly with broad subscription access, OpenAI has narrowed its scope and moved into the enterprise gate, and Google relies on a unique model that no one else has. How it plays out could be an early signal for how AI vendors will compete in other professional fields such as law, finance, and engineering in the future.
Claude Science is available in beta to anyone with a Pro, Max, Team, or Enterprise subscription. Anthropic also cited Novo Nordisk and Allen Research Institute as customer examples, suggesting the pharmaceutical company is already working with multiple AI vendors.
Anthropic will also support up to 50 Claude Science projects and provide up to $30,000 in credits. “We are looking for postdoctoral and graduate projects that explore the boundaries of science across disciplines with an early focus across biomedical research. Applications are being accepted until July 15, 2026, and award notifications will be sent by July 31. Projects will run from September 1 to December 1, 2026.”
If you buy through links in our articles, we may earn a small commission. This does not affect editorial independence.
