Anthropic’s Claude Science bets on workflow, not a new model, to win over scientists

On Tuesday, Anthropic launched Claude Science, an innovative AI workbench designed specifically for scientists. This platform aims to streamline computational research by offering a unified environment, eliminating the need for users to navigate multiple databases and tools. Notably, Anthropic clarified that Claude Science is not an entirely new or enhanced AI model; instead, it operates on existing Claude models, including Claude Opus 4.8, which are accessible to all users without restrictions. The introduction of Claude Science builds on the momentum of Claude for Life Sciences, which was launched in October 2025. This earlier initiative enhanced the Claude chatbot's capabilities in life sciences tasks. Claude Science, however, takes this a step further by providing a dedicated workspace tailored for scientific inquiries. This launch, revealed during an AI for Science briefing, aligns with Anthropic's strategy to move beyond being merely a model provider, aiming to establish a comprehensive operating layer for specialized industries, much like Claude Code has done for software development. At the core of Claude Science is a principal AI assistant that functions as a project manager for researchers. It connects seamlessly to over 60 scientific databases and includes pre-built toolkits for various fields, such as genomics, protein structure, and chemistry. This main assistant can create specialized sub-assistants to distribute tasks effectively, resembling a project lead assigning duties to team experts. Additionally, a separate AI fact-checker is incorporated to verify citations and calculations prior to publication, addressing concerns over accuracy in AI-assisted writing. Anthropic emphasizes that Claude Science includes features to enhance reproducibility. For instance, the workbench can produce figures like 3D protein structures alongside the code used to create them. Each figure comes with comprehensive documentation, including the exact code, a layman-friendly description, and the complete message history of the project. This functionality allows scientists to make edits in simple language, prompting the AI to adjust the underlying code accordingly. Another significant advantage of Claude Science is its ability to operate on a lab's existing infrastructure, removing the need to send data to Anthropic's servers. Early adopters have already reported positive outcomes. Sean Whalen, a principal scientist at Gladstone Institutes, successfully developed a genome browser from scratch in just days using Claude Science. Similarly, Jérôme Lecoq, a neuroscientist at the Allen Institute, utilized the platform to create a multi-agent computational review pipeline. This launch comes amid increasing competition in the scientific AI landscape. In April, OpenAI introduced GPT-Rosalind, a model specifically tailored for biological reasoning, albeit with limited access to select enterprise customers. Meanwhile, Google DeepMind focuses on leveraging proprietary science models like AlphaFold and AlphaGenome, which differ significantly from Anthropic's broad access approach. As the market evolves, the contrasting strategies of Anthropic, OpenAI, and Google could signal future competition in specialized sectors such as law, finance, and engineering. Currently, Claude Science is available in beta to users subscribed to Pro, Max, Team, and Enterprise plans. Anthropic is also supporting a maximum of 50 Claude Science projects, offering up to $30,000 in credits for innovative research initiatives, with applications open until July 15, 2026.