Anthropic introduced Claude Science on Tuesday, an AI workbench aimed at facilitating computational research for scientists by consolidating various databases, pipelines, and tools into a single environment. The company clarified that Claude Science is “not a new AI model and not a more capable model for biology,” as it operates on existing Claude models, including Claude Opus 4.8, without special access or gating.
The launch builds on the earlier introduction of Claude for Life Sciences in October 2025, which enhanced the Claude chatbot’s capabilities for life sciences tasks. The announcement was made during an AI for Science briefing, reflecting Anthropic’s strategy to expand beyond just providing models to creating workflows tailored for specific industries.
Claude Science features a primary AI assistant that functions as a project manager, linking to over 60 scientific databases. It includes prebuilt toolkits for disciplines such as genomics, protein structure, and chemistry. This assistant can create sub-assistants to delegate tasks effectively, allowing users to transfer work to customized expert assistants built for specific research needs. A separate fact-checker AI is used to verify citations and calculations before publication.
This verification process addresses concerns about fabricated citations in AI-generated writing, although it still uses the same underlying model for checking. To promote reproducibility, Claude Science allows users to generate figures alongside their corresponding code, providing details about the environment used and explanations of how the figures were created.
The platform is designed to save researchers time by enabling computations on their own infrastructure as opposed to relying on Anthropic’s servers. Early adopters include Jérôme Lecoq of the Allen Institute, who utilized Claude Science to develop a multi-agent computational review pipeline, and Stephen Francis’s team at UCSF, which expedited germline analysis of glioma significantly.
The launch comes shortly after OpenAI released GPT-Rosalind—a model tailored for biological reasoning—in April. OpenAI’s offering was gated to qualified enterprise customers, contrasting with Anthropic’s broader subscription model. Additionally, Google DeepMind approaches the market differently, providing proprietary foundational science models like AlphaFold.
Anthropic intends to support up to 50 Claude Science projects with grants of up to $30,000, specifically targeting postdoctoral and graduate research initiatives across multiple domains. Applications will be accepted until July 15, 2026, and projects are scheduled to run from September 1 to December 1, 2026.
