Anthropic launched Claude Science on Tuesday, an AI-powered workbench designed to give scientists a unified environment for conducting computational research. The platform aims to eliminate the inefficiency of switching between databases, pipelines, and tools throughout the research process.
Not a New Model, but a New Approach
Anthropic is clear that Claude Science is “not a new AI model and not a more capable model for biology. It runs the same Claude models already available to everyone today (including Claude Opus 4.8), with no special access and no gating.” The emphasis is on workflow integration rather than raw model power.
Building on Claude for Life Sciences
Claude Science builds on Anthropic’s October 2025 launch of Claude for Life Sciences, which enhanced the Claude chatbot’s performance in life sciences tasks. The new workbench provides a dedicated, structured environment for such specialized work, rather than just augmenting a general-purpose chatbot.
Vertical Strategy: Beyond Model Provider
Announced at an AI for Science briefing on Tuesday, the launch reflects Anthropic’s broader strategy to move beyond being just a model provider. The company aims to own the operating layer for specific industries—much like Claude Code has become for software development. Anthropic is increasingly betting its growth on vertical, workflow-level products rather than raw model capability, a shift that could reshape how it competes and prices against rivals like OpenAI.
How Claude Science Works
At the core of the platform is a main AI assistant that acts as a project manager for scientific research. It connects to more than 60 scientific databases and comes with pre-built toolkits for fields such as genomics, protein structure, and chemistry. This lead assistant can create sub-assistants to delegate specific tasks, similar to a project lead assigning work to specialists, or hand off complex work to custom “expert” assistants built by the user. A separate AI fact-checker then reviews citations and calculations before any output is finalized for publication.
This fact-checking capability is particularly relevant given the rise of AI-generated writing in science, which has led to fabricated citations and unverifiable statistics appearing in papers. However, since the same underlying model performs the verification, it remains an internal check rather than an independent source of truth.
Reproducibility and Time Savings
Claude Science includes features to ensure reproducibility. For example, the workbench can generate figures—such as 3D protein structures and chemistry diagrams—alongside the code that produced them. Each figure includes “the exact code and environment that produced it, a plain-language description of how it was created, and the full message history,” according to Anthropic. Scientists can also edit figures using natural language, prompting the AI agent to modify the underlying code accordingly.
The platform additionally saves time by allowing research to run on the lab’s own infrastructure, rather than transmitting sensitive data to Anthropic’s servers—a critical consideration for data security and compliance in scientific settings.
Early Adoption and Real-World Impact
Early users report significant productivity gains. Sean Whalen, a principal scientist in machine learning and functional genomics at Gladstone Institutes, used Claude Science to build a genome browser from scratch in just days, according to Anthropic. Similarly, Allen Institute neuroscientist Jérôme Lecoq leveraged the tool to create a multi-agent computational review pipeline, reducing what would have been years of human effort.
Competitive Context
The launch of Claude Science comes months after OpenAI took a different approach to the same problem. In April, OpenAI released GPT-Rosalind, a specialized model fine-tuned for biological reasoning. While OpenAI focused on model specialization, Anthropic is betting that workflow integration and tooling will prove more valuable for scientists—a bet that could define the next phase of competition in AI-powered research.”
via TechCrunch AI