The standalone AI workbench is positioned on the same tier as Anthropic's Claude Code coding assistant and publicly aims at the scientific workflows pioneered by Google DeepMind's AlphaFold protein folding ecosystem — though independent benchmarks and user evidence remain thin.
Anthropic on Tuesday launched Claude Science, its first flagship AI product built specifically for pharmaceutical researchers and biotech developers, positioning the tool on the same product tier as its mainstream Claude Code coding assistant and publicly aiming at the scientific workflows pioneered by Google DeepMind's AlphaFold protein-folding ecosystem. (MIT Technology Review, Anthropic)
The product is now generally available to all paid Claude subscribers. It is built to autonomously carry out multi-step research tasks from concise instructions and is wired to tools tuned for computational biology and drug development work. (Anthropic) The launch audience, as MIT Technology Review reports, was pharma executives, biotech founders, and bench researchers, the same constituency that has built serious scientific workflows around DeepMind's AlphaFold model and its successors. The bet, as TechCrunch frames it, is that integration and workflow beat raw model novelty in this category.
That is a deliberate strategy choice, not a capability ceiling. TechCrunch reports that Anthropic is not pitching a new foundation model. It is pitching a workflow layer that wraps existing Claude capabilities around the specific tools biologists and chemists already use. Elevation to the same product tier as Claude Code signals that Anthropic treats science workflows as a durable revenue category rather than a vertical demo, the same way Claude Code became the canonical surface for AI-assisted software work.
The contrast with last fall matters. In October 2025, Anthropic released "Claude for Life Sciences" as a set of plug-ins for the standard Claude product. MIT Technology Review describes Claude Science as a full-featured, standalone product distinct from those plug-ins, a sign that life-science users have graduated from add-on status to a dedicated surface. A companion research post, "Vibe physics: The AI grad student," extends the same thesis into physics research, with Anthropic positioning Claude as a collaborator capable of driving end-to-end investigations.
The self-application beat is real but narrow. Anthropic says it will use Claude Science internally to pursue drugs for rare, neglected diseases. The announcement names no specific disease, no timeline, and no external partner, and it does not yet constitute evidence of impact. It is a stated commitment and a marketing signal, not a validated research result.
The launch is mostly vendor-claimed. Independent benchmarks on real scientific task success rates are not yet public. Peer-reviewed results from third-party labs running Claude Science on drug-discovery programs are not in the record. As MIT Technology Review puts it, Anthropic wants to give the impression of being serious about AI-for-science. The product, the pricing, and the third-party pilots will tell us whether the impression holds.
What to watch next: independent benchmark numbers on biology task success, named pharma pilots that put Claude Science on a real drug program, and any movement on the rare-disease research commitment beyond the announcement stage.