Anthropic launched Claude Science, an AI workbench aimed at simplifying computational research for scientists by integrating multiple databases and tools into a single platform. The company clarified that Claude Science is not a new AI model but utilizes existing Claude models, including Claude Opus 4.8, available to all users without special access.
The workbench builds on the October 2025 launch of Claude for Life Sciences, which enhanced the Claude chatbot for life sciences tasks. The release was announced during an AI for Science briefing, emphasizing Anthropic’s strategy to provide vertical, workflow-level products instead of just standalone AI models.
Claude Science features a main AI assistant that acts as a project manager, linking to over 60 scientific databases. It includes prebuilt toolkits for specific domains such as genomics, protein structure, and chemistry. This assistant can also create sub-assistants for task delegation or custom expert assistants tailored for individual research projects. A separate fact-checker AI is designed to verify citations and calculations before publication.
Anthropic noted the importance of the fact-checking step as AI-assisted writing increasingly risks fabricating citations and unverifiable statistics. The fact-checking tool relies on the same underlying model as the main assistant, maintaining the need for rigorous validation.
The system enhances reproducibility by generating visual figures, including 3D protein structures, and supplying the corresponding code and descriptions for their creation. Users can save time by editing figures in plain language, prompting automatic adjustments to the underlying code.
Claude Science can operate on a lab’s infrastructure, allowing data processing locally and reducing the need to rely on Anthropic’s servers. Early adopters, including Jérôme Lecoq from the Allen Institute, reported significant time savings in their research workflows. Stephen Francis’s group at UCSF also found that Claude Science expedited their germline analysis of glioma significantly.
This launch follows OpenAI’s introduction of GPT-Rosalind in April, a model focused on biological reasoning that is gated for select enterprise customers. The contrasting distribution strategies highlight how Anthropic aims to offer broad subscription access, while OpenAI restricts access with qualification processes.
Competing directly in this field is Google DeepMind, which has proprietary models such as AlphaFold and its Gemini for Science platform that integrates multiple life science databases.
Claude Science is now available in beta for users with Pro, Max, Team, and Enterprise subscriptions. Anthropic is supporting up to 50 research projects with funding of up to $30,000 each, focusing on postdoctoral and graduate work in biomedical research. Applications for project support are accepted until July 15, 2026, with award notifications by July 31, and projects scheduled to run from September 1 to December 1, 2026.
