SandboxAQ integrates physics-grounded AI into Claude

SandboxAQ is integrating its physics-grounded scientific AI models into Claude, the AI platform provided by Anthropic. The integration allows users to access complex scientific simulations through a conversational interface rather than relying on specialized computing infrastructure. Unlike standard text-based large language models (LLMs) that rely on patterns in text, SandboxAQ’s models are built on physical world rules and equations. These models, known as Large quantitative models (LQMs), are engineered to run quantum chemistry calculations and simulate both molecular dynamics and microkinetics, which is the study of how chemical reactions unfold at the molecular level. This simulation capability allows researchers to see how candidate molecules are likely to behave before conducting physical laboratory experiments.

The company, which was founded roughly five years ago and counts Eric Schmidt as its chairman, has raised more than $950 million from investors. With this integration, SandboxAQ is targeting what it describes as the quantitative economy—a sector spanning biopharma, financial services, energy, and advanced materials that the company states is a $50+ trillion sector. The company notes that these LQMs are trained on real-world lab data and scientific equations to serve this market, which represents one of several business lines the company has built out.

This focus on conversational accessibility distinguishes SandboxAQ from competitors in the AI model space, such as Chai Discovery and Isomorphic Labs, which have focused primarily on the underlying science. Previously, users of SandboxAQ’s models had to provide their own digital infrastructure to run simulations. By integrating with Claude, the company aims to simplify this process for computational scientists, research scientists, and experimentalists working at large pharmaceutical or industrial firms who are searching for new materials that can become marketable products. Nadia Harhen, SandboxAQ’s general manager of AI simulation, emphasized this shift in accessibility: “For the first time, we have a frontier [quantitative] model on a frontier LLM that someone can access in natural language.” Harhen noted that customers often turn to SandboxAQ after trying other software that failed to yield positive real-world results for highly complex problems, explaining that their issues often do not work or do not yield positive results when translating to the real world.

Why it matters

SandboxAQ is integrating its scientific AI models into Claude to remove the need for specialized computing infrastructure, aiming to make drug discovery and materials science tools accessible via a conversational interface. By lowering the technical barrier to running complex physical simulations, the partnership could expand the use of scientific AI across the broader quantitative economy.