Anthropic launches Claude Science drug-discovery workbench, integrates NVIDIA BioNeMo

Anthropic launched Claude Science, an AI workbench for scientists that enables tasks like antibiotic design and vaccine prediction through natural-language prompts. In a coordinated move, Anthropic integrated Claude Science with NVIDIA BioNeMo — including its libraries, models, and NIM microservices — allowing Claude Science agents to direct scientific workflows in natural language while accessing NVIDIA compute resources.

The integration targets the fast-growing market for AI tools in drug discovery, genomics, and molecular design, leveraging BioNeMo's established presence in pharmaceutical research. By pairing Claude's reasoning and agentic planning with BioNeMo's domain-specific models, the aim is to let researchers orchestrate complex computational-biology pipelines — from target identification to molecular design — without deep programming expertise.

The launch positions Anthropic directly against OpenAI's scientific push, which this week included GeneBench-Pro, a benchmark for judgment-heavy computational biology on which GPT-5.6 Sol posts strong results. Both labs are racing to prove frontier models can materially accelerate real research, not just answer questions — a higher bar that computational biology, with its messy data and expert judgment calls, tests well.

Supporting the case, practitioners cited a documented instance of Claude materially aiding a Nobel laureate's team in solving a decades-old physics jamming problem, offered as real evidence of AI-assisted formal research. Claude Science arrives alongside Anthropic's Sonnet 5 launch and the lifted export controls, part of a busy product week ahead of its IPO. Watch for peer-reviewed validation of Claude Science outputs and pharma-partner adoption of the BioNeMo integration.