Anthropic launches Claude Science, a workflow platform for researchers that doesn't require a new AI model

Anthropic announced Claude Science on Tuesday, a dedicated computational workbench designed to streamline how scientists conduct research by consolidating disparate databases, data pipelines, and analytical tools into a single environment.

The company has been explicit about what Claude Science is not: it does not introduce a new AI model or a more capable variant optimized for biology. Instead, it deploys the same Claude models—including Claude Opus 4.8—that are already publicly available, with no exclusive capabilities or access restrictions.

Building on life sciences push#

The launch extends momentum from Anthropic's October 2025 introduction of Claude for Life Sciences, which enhanced Claude's proficiency in biology-related tasks. Claude Science provides a dedicated operating environment to harness that capability.

The platform announcement, made at a Tuesday briefing on AI for science applications, reflects Anthropic's strategic direction: positioning itself beyond a model vendor and toward owning the workflow layer for specialized industries. This mirrors how Claude Code has emerged as the primary development environment for software engineers. The company is increasingly betting its future on vertical, industry-specific products rather than relying solely on raw model improvements.

How the workspace operates#

A primary AI assistant functions as a research project manager, connecting to over 60 scientific databases and bundling pre-built tools tailored to specific disciplines—genomics, protein structure analysis, chemical research, and others. The assistant can spawn secondary agents to distribute tasks, similar to a lead researcher delegating work to team members, or activate custom-built "expert" assistants aligned with individual research agendas. A separate fact-checking component then validates citations and computations before publication.

That validation step responds to a real concern: as AI-generated content proliferates, fabricated references and unsupported statistics increasingly appear in academic papers. However, Anthropic acknowledges this checker relies on the same underlying model rather than serving as an independent verification source.

Reproducibility and data control#

Claude Science emphasizes research reproducibility in several ways. It generates figures—such as 3D protein models or chemical diagrams—alongside the executable code that produced them. Every visualization includes the underlying code, plain-language documentation of its creation process, and the full conversation history. Scientists can also revise figures using natural language, and the system automatically adapts the corresponding code.