Anthropic’s Claude Science: The AI Lab That Wants to Cure Diseases and Dominate Drug Discovery
Anthropic has launched Claude Science, a flagship agentic AI product designed for scientific research and drug discovery, positioning itself to challenge Google DeepMind's decade-long dominance in AI for science. The product is now available to paid Claude subscribers and Anthropic plans to use it for its own drug discovery research into neglected diseases.
Anthropic Enters the Lab: What Claude Science Really Means
For years, when the conversation turned to AI and scientific discovery, one name dominated the room: Google DeepMind. Its AlphaFold model reshaped protein biology so profoundly that CEO Demis Hassabis and researcher John Jumper won a Nobel Prize in chemistry for it. But the frontier of AI moves fast, and on Tuesday, Anthropic made a move that signals it is ready to claim scientific AI as its own territory — announcing Claude Science, a major new flagship product aimed squarely at researchers, biotech founders, and pharmaceutical companies.
As detailed in a report by MIT Technology Review (https://www.technologyreview.com/2026/06/30/1139987/claude-science-is-anthropics-newest-flagship-product/), Claude Science is designed to support scientific research with the same depth and autonomy that Claude Code brings to software engineering. It is now available to all paid Claude subscribers, and its elevation to flagship status — sitting alongside Claude Code and Claude Cowork — tells you everything about how seriously Anthropic is treating this bet.
What Claude Science Actually Does
At its core, Claude Science is an agentic AI system. Like Claude Code, it can take high-level, concise instructions and carry out meaningful work autonomously — without the user needing to micromanage every step. But it goes significantly further than a generic research assistant.
The product is purpose-built for the workflows of molecular and cellular biology, computational biology, and drug development. It can interface with specialised tools used in genetics, chemistry, and protein biology — domains that sit at the intersection of complex data analysis and scientific discovery. Crucially, it does not just write code; it helps scientists actually run that code on powerful computing clusters, which are essential for large-scale biological simulations but notoriously difficult to manage without dedicated IT support. For researchers at smaller institutions or startups, that capability alone could be transformative.
Reproducibility is another priority baked into Claude Science’s design. In scientific research, the ability to trace back any result, figure, or conclusion to its original source is not a nice-to-have — it is a fundamental requirement of credible science. Claude Science prioritises this, allowing researchers to audit outputs and verify their validity at every step.
A Live Demo That Made the Case
At the launch event attended by pharmaceutical executives and biotech founders, Alexander Tarashansky, who led the development of Claude Science, gave a live demonstration of the system autonomously identifying new drug candidates for phenylketonuria — a rare genetic disease. That demonstration was not chosen arbitrarily. It illustrated the product’s most commercially compelling use case: accelerating early-stage drug discovery in an industry where the cost and time to bring a drug to market are staggering.
Why This Matters: The DeepMind Comparison
For the past decade, DeepMind set the standard for what AI could do in science. AlphaFold, contributions to meteorology, materials science — the list is long and genuinely impressive. But as MIT Technology Review notes, the fast-advancing frontier of AI progress has recently left DeepMind struggling to keep pace, particularly in the coding domain that has become the most commercially valuable application of large language models.
Anthropicis now well positioned to step into that vacuum. Anthropic CEO Dario Amodei holds a PhD in computational neuroscience — a scientific credential that Sam Altman at OpenAI does not share, and that aligns Anthropic’s leadership culture more closely with the research community it is trying to serve. Many working scientists are already using Claude Code as part of their daily workflows, precisely because modern research involves substantial programming but not every researcher is a software engineer.
The signal that perhaps matters most came earlier in June, when John Jumper — the DeepMind researcher who shared the Nobel Prize with Hassabis — announced he was leaving DeepMind for Anthropic. That kind of talent migration speaks louder than any press release.
Graduate Student or Research Partner? Setting Expectations Right
One of the most useful data points in understanding Claude Science’s current capabilities comes from Harvard physicist Matthew Schwartz. In a blog post on Anthropic’s website, Schwartz estimated — based on his own work with Claude Code and other Anthropic tools — that the company’s Opus 4.5 model performs at roughly the level of a second-year graduate student when executing scientific projects.
That framing is instructive. A second-year graduate student is not a Nobel laureate. They cannot independently design and execute a groundbreaking research programme. But they can run experiments, analyse data, write code, troubleshoot protocols, and produce work that meaningfully advances a project under the guidance of a senior researcher. Applied across thousands of research groups simultaneously, that capability represents a genuine acceleration of scientific output.
Eric Kauderer-Abrams, Anthropic’s head of life sciences, was clear that Claude Science is not meant to replace Claude Code or Claude Cowork in existing workflows. Rather, it builds on what researchers already find valuable and adds the domain-specific tools and integrations that pure coding agents lack.
Anthropic’s Own Drug Pipeline: Mission or Business Strategy?
Perhaps the most striking element of Tuesday’s announcement is that Anthropic itself plans to use Claude Science to pursue research into drug candidates for neglected diseases — conditions that affect millions of people globally but attract little commercial investment because the patients cannot afford expensive treatments.
This move has a dual purpose. On one level, it is a genuine expression of Anthropic’s stated mission to develop AI for humanity’s long-term well-being — a phrase Kauderer-Abrams echoed when he called life sciences “by far the greatest opportunity” to fulfil that mission. On another level, it gives Anthropic a real-world testbed to understand how Claude Science performs under actual research conditions, producing feedback that will make the product better over time.
The Commercial Reality
There is also a frank commercial logic at play here. Academic research institutions, while enthusiastic early adopters of AI tools, operate on tight budgets. Pharmaceutical companies do not. With Anthropic reportedly approaching its first profitable quarter and an IPO anticipated later in 2026, landing major contracts with large pharma firms would provide a durable, high-value revenue stream — one that is far less vulnerable to the commoditisation pressures squeezing the broader LLM market.
In India, where domestic pharmaceutical manufacturing is a significant industry and where biotech startups are attracting growing investment, Claude Science could find receptive users relatively quickly. Access pricing for paid Claude subscriptions — which would determine what Indian researchers and companies pay — will be worth watching as adoption builds.
The Bigger Picture: A New Era of AI-Assisted Science
What Anthropic is attempting with Claude Science is not just a product launch. It is a repositioning of where the company sits in the AI landscape — from a safety-focused LLM provider to an active participant in the future of biological research. The decision to pursue its own drug pipeline, to hire a Nobel-adjacent researcher, and to give Claude Science flagship status all point in the same direction.
DeepMind built its scientific reputation over a decade through foundational research. Anthropic is trying to compress that timeline by combining frontier model capability with domain-specific tooling and a commercial urgency that academic labs rarely feel. Whether it succeeds will depend on whether Claude Science can move from impressive demos to reproducible, publishable, therapeutically relevant results.
But for pharmaceutical executives and biotech founders — the exact audience sitting in the room on Tuesday — the pitch is already coherent: here is an AI system that can help you find drug candidates faster, run the compute you need without the infrastructure headaches, and trace every result back to its source. That is a compelling offer, regardless of how the broader AI competition plays out.
