SandboxAQ

SandboxAQ is a quantitative-AI company applying Large Quantitative Models (LQMs) — simulation-grade physics and mathematics models that complement large language models — to drug discovery, materials science, cybersecurity, and quantitative finance. The company was spun out of Alphabet (Google's parent) in 2022 and is led by CEO Jack Hidary, with backers including Eric Schmidt, Marc Benioff, Salesforce Ventures, and T. Rowe Price across multiple late-stage rounds. Headquarters is in Palo Alto, California.

SandboxAQ's defining technical bet is that language models alone are insufficient for scientific work in chemistry, physics, and biology — quantum chemistry, molecular dynamics, and microkinetics require simulation-grade numerical models that LLMs cannot replicate by token prediction. The Large Quantitative Models are purpose-built physics and chemistry simulators that run alongside LLMs, exposing chemistry-grade compute through software interfaces that scientists can drive without standing up their own high-performance-computing infrastructure.

In May 2026, SandboxAQ shipped its flagship integration with Claude, exposing LQMs for quantum chemistry, molecular dynamics, and microkinetics directly inside Anthropic's conversational interface. Computational scientists and research scientists at pharmaceutical and materials companies can now query simulation-grade physics models in natural language without provisioning their own digital infrastructure. SandboxAQ's general manager of AI simulation, Nadia Harhen, framed the launch as "the first time we have a frontier quantitative model on a frontier large language model that someone can access in natural language." The integration mechanism — whether through MCP, a managed API, or an embedded extension — has not yet been detailed by Anthropic.

The Claude integration positions SandboxAQ at the intersection of two adoption curves: the rapid spread of LLM interfaces inside enterprise R&D workflows, and the longer-running shift toward AI-assisted drug discovery and materials science. By selling LQM access against existing Claude seats rather than as a separate vertical platform, SandboxAQ removes the procurement and onboarding overhead that has historically slowed simulation-grade tooling adoption inside large pharmaceutical and chemical research organizations. Beyond pharmaceuticals, SandboxAQ's portfolio also covers post-quantum cryptography (AQtive Guard) and quantitative finance — areas where the same simulation-grade modeling approach applies.