Anthropic Unveils Claude Science to Accelerate Drug Discovery

Anthropic announced Claude Science at a gathering of pharmaceutical leaders, biotech founders, and academic researchers on Tuesday. The new model is positioned as the scientific counterpart to Claude Code, which already powers software‑engineering workflows. With concise, high‑level prompts, Claude Science can execute complex tasks—such as running molecular‑dynamics simulations or querying protein‑structure databases—by calling specialized tools built into the system.

The announcement followed a 2025 release of “Claude for Life Sciences,” a set of plug‑ins that let Claude access scientific software. Claude Science, however, is a full‑featured, standalone product that integrates those plug‑ins into a single, user‑friendly interface. Anthropic’s head of life sciences, Eric Kauderer‑Abrams, said the move signals the company’s commitment to “AI that serves humanity’s long‑term well‑being.”

In practice, a researcher can hand Claude Science a brief description of a target protein and ask it to suggest potential inhibitors. The model then queries public databases, runs docking simulations, and returns ranked candidates—all without manual intervention. Early demos showed the system identifying a promising scaffold for a rare‑disease enzyme in under ten minutes, a task that would normally require a team of chemists and computational biologists.

Anthropic plans to use Claude Science internally to explore drugs for neglected diseases, a strategy that mirrors DeepMind’s AlphaFold work on protein folding. DeepMind’s 2023 breakthrough earned a Nobel Prize in chemistry and set a benchmark for AI in biology. By contrast, Anthropic’s approach focuses on end‑to‑end pipelines that combine data retrieval, simulation, and synthesis planning.

The launch also coincides with a broader shift toward “workflow‑friendly” AI. Mapify’s recent guide notes that models like GPT‑5.4 and Gemini 3 Pro are gaining traction for their ability to chain reasoning, coding, and tool use. Claude Science joins this cohort as a specialized tool that can be deployed on NVIDIA‑accelerated infrastructure, a fact highlighted by NVIDIA’s own AI‑model catalog, which lists optimized inference engines for LLMs.

While the product is technically impressive, it is not without limitations. Claude Science’s tool‑calling relies on pre‑approved APIs, so it cannot yet access proprietary or highly regulated databases without additional integration work. Moreover, the model’s performance on highly novel chemical spaces remains untested; early benchmarks show accuracy comparable to existing commercial platforms but lag in discovering truly novel chemotypes.

Anthropic’s decision to elevate Claude Science to flagship status also reflects market dynamics. PricePerToken’s recent model‑release roundup notes a surge in specialized LLMs, with competitors like OpenAI’s GPT‑5.4 and Google’s Gemini 3 Pro vying for the same niche. By offering a dedicated biology tool, Anthropic differentiates itself from general‑purpose models that require extensive fine‑tuning.

The broader AI‑for‑science ecosystem is tightening its focus on bias mitigation. The Vector Institute’s UnBias‑Plus, released today, provides a free, open‑source tool that flags and rewrites biased language in scientific writing. Integrating such bias‑detection layers into Claude Science could help ensure that generated hypotheses do not inadvertently propagate existing inequities in drug discovery.

For practitioners, the key takeaway is that Claude Science is now a paid‑subscription product, meaning that researchers with existing Claude access can immediately experiment with it. Anthropic’s pricing model mirrors that of other LLM providers, offering a pay‑as‑you‑go tier for tool‑calling usage.

Looking ahead, the real test will be whether Claude Science can deliver clinically relevant leads at scale. If it can, the model could shift the drug‑discovery pipeline from months to weeks, a transformation that would reverberate across the life‑science industry.

FAQ

Q: Is Claude Science available to non‑paid users?
A: No, it is currently limited to paid Claude subscribers.

Q: What kinds of tools can Claude Science call?
A: The system supports a range of bioinformatics and cheminformatics APIs, including protein‑structure databases, molecular‑dynamics engines, and synthetic‑route planners.

Q: How does Claude Science compare to AlphaFold?
A: AlphaFold predicts protein structures; Claude Science orchestrates multi‑step research workflows, from data retrieval to candidate generation.

Q: Will it handle proprietary datasets?
A: Integration with private APIs is possible but requires custom configuration.