Anthropic targets research labs with on-premises Claude Science workbench
Anthropic has launched a workflow-focused AI research environment that runs on local lab infrastructure, setting up a three-way battle for European scientific computing budgets.
Anthropic introduced Claude Science on Tuesday, a dedicated AI workbench designed to let scientists run computational research without switching between separate databases and tools.
The product does not use a new or specialized AI model. It applies Anthropic’s existing Claude models, including Claude Opus 4.8, to a unified workspace. This represents a strategic shift by the company toward owning the workflow layer for specific industries rather than competing solely on raw model capability.
A central AI assistant manages projects and connects to more than 60 scientific databases, delegating tasks to custom sub-assistants. Crucially for European research institutions, the workbench can run entirely on a laboratory’s own infrastructure rather than sending data to Anthropic’s servers.
The system includes a separate AI fact-checker to verify citations and calculations before publication, though Anthropic concedes this is still the same underlying model checking itself. It also ensures reproducibility by tying generated 3D protein structures and chemistry drawings to the exact underlying code, plain-language descriptions, and full message history.
Anthropic’s approach contrasts sharply with its rivals. OpenAI recently released GPT-Rosalind, a specialized biological reasoning model restricted to qualified US enterprise customers. Google DeepMind is taking yet another path, bundling its proprietary foundational models like AlphaFold and AlphaGenome into its Gemini for Science platform.
These three distinct distribution strategies—broad subscription access versus enterprise gating versus proprietary model bundling—are now clashing over the scientific research market. The outcome will likely signal how AI vendors approach other specialized European verticals like finance, law, and engineering.
Pharma companies are already spreading their bets. Anthropic named Novo Nordisk and the Allen Institute as early case studies, indicating that major European drugmakers are actively testing multiple AI vendors simultaneously. Early adopters have already put the tool to practical use, with Stephen Francis’s group at the UCSF Brain Tumor Center using it to reduce the time required for comprehensive germline analysis of glioma.
Claude Science is available in beta for Anthropic’s Pro, Max, Team, and Enterprise subscribers. The company is also offering up to $30,000 in credits for 50 postdoctoral and graduate projects, stating it is “looking for postdoctoral and graduate projects that span domains and explore the boundaries of science, with an early focus on fields across biomedical research.”
