Anthropic released Claude Science on Tuesday, a workbench that gives researchers a single computational environment tied to more than 60 scientific databases - a move that turns the AI assistant into an operating layer for lab work and escalates a three-way fight with OpenAI and Google DeepMind over the scientific research market.
How Claude Science works
One main AI assistant functions as a project manager, connecting to prebuilt toolkits for genomics, protein structure, and chemistry. That assistant can create sub-assistants to split tasks or hand work to custom "expert" assistants the user has built for their own research. A separate fact-checker AI then re-checks citations and calculations before anything is readied for publication.
Anthropic said the workbench "not a new AI model and not a more capable model for biology. It runs the same Claude models already available to everyone today (including Claude Opus 4.8), with no special access and no gating." It builds on the company's October 2025 Claude for Life Sciences release, which improved the chatbot's life sciences performance.
Fact-checking and reproducibility
As AI-assisted writing makes fabricated citations harder to catch, the fact-check step is designed to surface errors before they reach a paper. The checker still relies on the same underlying model, not an independent source, but Anthropic has added other reproducibility safeguards. The workbench generates figures like 3D protein structures and chemistry drawings alongside the exact code and environment that produced them, plus a plain-language description and full message history. Scientists can edit figures by giving instructions in plain language, prompting the assistant to update its own underlying code.
Allen Institute neuroscientist Jérôme Lecoq used the tool to build a multi-agent computational review pipeline. Stephen Francis's group at the UCSF Brain Tumor Center cut the time required for germline analysis of glioma to a fraction of what it previously took, with results independently validated.
Competing strategies in AI for science
The launch comes two months after OpenAI released GPT-Rosalind, a specialized model fine-tuned for biological reasoning and gated behind a qualification and safety review for select U.S. enterprise customers. Google DeepMind, meanwhile, controls foundational science models like AlphaFold and AlphaGenome and bundles them with more than 30 databases inside its Gemini for Science platform. Those three distribution approaches - Anthropic's broad subscription access, OpenAI's gated enterprise preview, and Google's proprietary model ecosystem - are now competing for the same labs and pharmaceutical research budgets.
The competition highlights a growing demand for AI for Science & Research tools that go beyond raw model capability and embed directly into scientific workflows.
Why this matters for scientists and researchers
Claude Science is available in beta to anyone on Pro, Max, Team, and Enterprise subscriptions. The workbench runs on a lab's own infrastructure setup, keeping data off Anthropic's servers. Novo Nordisk and the Allen Institute are already named as early customers, suggesting that pharma organizations are testing multiple AI vendors.
Anthropic will also support up to 50 Claude Science projects with up to $30,000 in credits. The call is open to postdoctoral and graduate projects across biomedical research; applications close July 15, 2026, with awards announced July 31. For researchers, the bigger signal is structural: the three largest AI firms are each building vertically integrated science platforms, and which ecosystem a lab commits to could shape data security, tool access, and reproducibility standards for years.
Your membership also unlocks:
