Anthropic puts Claude to work in wet labs and pharma

Most scientists use AI the way the rest of us do: summarizing papers, cleaning up Python scripts, or drafting mundane emails.

Anthropic wants to change that. The company just introduced Claude for Life Sciences, a tailored suite aimed at turning its models into full-fledged lab partners capable of supporting the entire drug discovery and development lifecycle, from literature reviews to regulatory filings.

Under the hood sits Claude Sonnet 4.5. On Protocol QA, a benchmark testing how well a model understands and executes laboratory protocols, Sonnet 4.5 scored 0.83, beating Sonnet 4’s 0.74 and topping the human expert baseline of 0.79. It also posted solid jumps on BixBench, which tracks bioinformatics reasoning.

Wiring Claude directly into wet lab tools

High benchmark scores are nice, but wet labs run on specialized software, not generic chat boxes.

Anthropic rolled out a fleet of dedicated connectors to bridge the gap. Claude can now plug directly into Benchling, answering scientist queries while citing actual experiment notebooks. It integrates with BioRender for scientific diagrams, PubMed and Wiley’s Scholar Gateway for peer-reviewed literature, Synapse.org for data sharing, and 10x Genomics to run single-cell and spatial analysis using plain English.

To lock down scientific workflows, Anthropic is also leveraging its recently launched Agent Skills framework. The company built an out-of-the-box skill called single-cell-rna-qc, which executes standardized quality control and filtering on single-cell RNA sequencing data following scverse best practices. Labs can also package their own custom internal protocols into reusable skills.

The offering is live on Claude.com and the AWS Marketplace, with Google Cloud Marketplace coming soon. Anthropic is also offering free API credits through its AI for Science program to hook academic researchers early, backed by enterprise rollout muscle from Deloitte, Accenture, PwC, and KPMG.

Why it matters

Biology and drug discovery are arguably the most lucrative yet punishing battlegrounds in artificial intelligence. What makes Anthropic’s move clever is that it isn't trying to replace established biopharma platforms. Instead, it is turning Claude into universal connective tissue across Benchling, 10x Genomics, and cloud data warehouses like Databricks and Snowflake.

For bench researchers, this means less time wrestling with data formatting and syntax, and more time interrogating results in natural language. If you run a life sciences lab, using Claude to parse massive literature pools, draft SOPs, and run routine quality control scripts is well worth exploring today.

Still, skepticism is warranted. Hallucinations in life sciences aren't harmless typos; they can derail expensive clinical pipelines or corrupt regulatory compliance documents. Even with Sonnet 4.5 beating human baselines on paper, real-world biology is notoriously messy, and human verification remains indispensable.

Strategically, this signals where the frontier model race is heading. While OpenAI frequently targets broad knowledge work and developer tooling, Anthropic is systematically staking claims on specialized, high-budget enterprise verticals. By pairing free academic research grants with heavy-hitting consulting partners, Anthropic is quietly locking down the future biopharma pipeline before competitors can catch up.