Claude Science: Anthropic launches an AI research workbench for scientists

Anthropic launched Claude Science, a dedicated research platform for scientists that integrates more than 60 scientific databases, domain-specific toolkits, and a multi-agent workflow with built-in citation verification. The platform runs on existing Claude models (including Claude Opus 4.8) and is available in beta for Pro, Max, Team, and Enterprise subscribers.

Key takeaways

Not a new model — Claude Science runs on the same Claude models already available to everyone
Integration with 60+ scientific databases — prebuilt toolkits for genomics, proteomics, and chemistry
Multi-agent architecture: a main assistant delegates tasks to specialized sub-agents
Built-in citation and calculation verification agent before publication
Grant program: up to $30,000 in credits for 50 postdoctoral and graduate projects (applications open until July 15, 2026)

A platform, not a model

Anthropic is explicit: "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." There is no special access to biological data and no additional fine-tuning for life sciences.

Claude Science is a research environment — an interface that consolidates a fragmented ecosystem of scientific tools into a single workspace. Scientists no longer have to switch between databases, computational pipelines, and text editors. It builds on Claude for Life Sciences, which Anthropic launched in October 2025 as a chatbot augmentation for life sciences tasks.

Architecture: project manager and specialist agents

The central unit is the main assistant, which acts as a project manager. It can create sub-agents for specific tasks (searching GenBank, calculating protein structures, generating charts) or hand off work to a user-configured "expert" assistant built for a specific research project.

All generated figures — 3D protein structures, chemical diagrams — are stored alongside the code that produced them, a plain-language description of the method, and the full conversation history. Editing a figure means issuing a natural-language instruction: the assistant modifies its own underlying code. This addresses one of the most common problems in AI-assisted scientific writing — the inability to reproduce a chart.

A fourth component is a fact-checking agent that verifies citations and calculations before anything goes to publication. Anthropic acknowledges the limitation: the same model checks its own output, rather than an independent source of truth.

Competitive context

Claude Science launches as the AI-for-science market grows more crowded. OpenAI released GPT-Rosalind in April 2026 as a specialized biological reasoning model — but access is gated to qualified US enterprise customers including Amgen, Allen Institute, Moderna, Thermo Fisher, and Novo Nordisk. Google DeepMind takes a different route: it owns foundational science models (AlphaFold, AlphaGenome) and bundles them with 30+ life science databases in its Gemini for Science platform.

The strategic difference is clear. Anthropic bets on broad access (subscription, no qualification gate), OpenAI bets on a narrow enterprise gate, and Google bets on proprietary science models no one else has.

Why it matters

Claude Science is Anthropic's wager that in specialized AI market segments, the best workflow integration matters more than the most powerful model. Scientists spend hours jumping between databases and pipelines. A single interface that consolidates this may be more valuable than a higher benchmark score.

The platform signals the direction Anthropic wants to grow: from model provider to vertical software provider. Claude Code became the operating layer for software engineers. Claude Science aims for the same position in scientific research. If it works, Anthropic will defend its margin through integration — which is harder to copy than raw model capability.

Citation verification addresses a real problem: AI-assisted scientific writing produces fabricated references with concerning regularity. Even partial mitigation gives the platform measurable added value.

What's next

Beta available from June 30, 2026 for Pro, Max, Team, and Enterprise subscribers. No general availability timeline announced
Grant application deadline: July 15, 2026. Award notifications by July 31, projects run September 1 to December 1, 2026
Anthropic plans to expand beyond biology — Claude Science launches with a biomedical research focus but has not specified a timeline for other scientific domains