SandboxAQ has made its first Large Quantitative Model integration available through an LLM interface, starting with AQCat Adsorption Spin, a model aimed at catalyst discovery workflows.

What changed

In a May 18 post, SandboxAQ said researchers can now reach some of its physics-based models through natural-language prompts rather than working directly with specialized simulation tooling. The company describes its Large Quantitative Models, or LQMs, as systems trained on physics-grounded data and used inside automated scientific workflows.

The first live integration is AQCat Adsorption Spin, which SandboxAQ says helps estimate adsorption energy, an early screening step in heterogeneous catalyst discovery. Its material-discovery page adds that the integration is built on MCP and can be accessed through compatible MCP clients. SandboxAQ is also taking waitlist signups for access and says additional material-discovery integrations are planned.

Why it matters

The conservative read is that this is an access-layer release, not a new foundation model. SandboxAQ is trying to put domain-specific quantitative models behind the same conversational interfaces researchers already use for general AI work.

That could matter if it reduces the friction between scientific question, model routing, and candidate screening. For now, the useful signal is narrow but concrete: a specialized catalyst model is moving into an LLM-driven workflow, with human researchers still deciding which results are worth taking into modeling or lab work.