Anthropic on Tuesday unveiled Claude Science, an AI workbench intended to give researchers a single environment for computational work so they don’t have to switch among databases, pipelines and tools.
Anthropic emphasized that Claude Science is “not a new AI model and not a more capable model for biology.” The workbench runs the same Claude models already publicly available (including Claude Opus 4.8), without special access or gating.
Platform roots and main capabilities
Claude Science builds on Anthropic’s October 2025 rollout of Claude for Life Sciences, which improved the Claude chatbot’s performance on life‑sciences tasks. Claude Science is positioned as a dedicated workspace for those activities.
A primary AI assistant acts like a project manager: it connects to more than 60 scientific databases and offers prebuilt toolkits for domains such as genomics, protein structure and chemistry. That assistant can spawn sub‑assistants to divide labor or hand tasks off to custom “expert” assistants that users create for their own research. A separate fact‑checker AI then re‑validates citations and calculations before results are prepared for publication. Anthropic notes, however, that this validation step still uses the same underlying model rather than an independent ground truth source.
Reproducibility and integrated artifacts
Anthropic says Claude Science supports reproducibility by saving figures—like 3D protein structures and chemical drawer images—alongside the exact code and environment that produced them. Each figure includes a plain‑language description of how it was created and the full message history. The platform also lets researchers edit figures using natural language prompts, which then cause the agent to modify its own underlying code.
Data locality option
Claude Science can run on a lab’s local infrastructure instead of sending data back to Anthropic’s servers, which can save time and address data‑security preferences.
Early adopters and use cases
Early users report practical gains. Jérôme Lecoq of the Allen Institute used Claude Science to build a multi‑agent computational review pipeline. Stephen Francis’s team at the UCSF Brain Tumor Center used Claude Science to compress the time needed for comprehensive germline analysis of glioma to a fraction of previous timelines, with results independently validated.
Competing approaches from OpenAI and Google DeepMind
The Claude Science launch follows other moves in the same space. In April, OpenAI released GPT‑Rosalind, a model fine‑tuned for biological reasoning that debuted as a research preview limited to qualified enterprise customers in the U.S., gated by qualification and safety review; early partners included Amgen, Allen Institute, Moderna, Thermo Fisher and Novo Nordisk.
Google DeepMind takes a different tack by owning foundational science models such as AlphaFold and AlphaGenome, which others can only call as tools. DeepMind’s Gemini for Science bundles those models and more than 30 life‑science databases into a single skill set.
Those three strategies amount to distinct distribution plays in the scientific research market: Anthropic is pursuing broad subscription access, OpenAI is offering a narrow, enterprise‑gated preview, and Google DeepMind relies on proprietary, in‑house models.
Availability and grant program
Claude Science is available in beta to anyone on Pro, Max, Team or Enterprise subscriptions. Anthropic named Novo Nordisk and the Allen Institute as customer case studies.
Anthropic will support up to 50 Claude Science projects with up to $30,000 in credits per project. The company is targeting postdoctoral and graduate projects that span domains, with an early emphasis on biomedical research. Applications are open through July 15, 2026; award notifications will be sent by July 31, and selected projects will run from September 1 to December 1, 2026.
