SandboxAQ brings its drug discovery models to Claude – no PhD in computing required

Drug discovery is one of the most costly failures in modern industry. Finding a single viable molecule could take a decade and cost billions, and most candidates still can’t make it happen. A generation of AI startups have promised to fix this, with most of them making the problem less painful for researchers, who already have enough technical expertise to use the tools.

But SandboxAQ believes the bottleneck is not in the models. It’s the interface.

The company has teamed up with Anthropic to integrate its scientific AI models directly into Claude, putting powerful drug discovery and materials science tools behind a conversational interface that doesn’t require specialized computing infrastructure to use.

SandboxAQ was founded nearly five years ago as a spinoff of Alphabet, and includes former Google CEO Eric Schmidt as its president. The company, which has raised more than $950 million from investors, has built a number of different business lines, including a cybersecurity business.

However, one of the unique things that SandboxAQ does is produce large quantitative models, or LQMs. These particular models are “physics-based,” meaning they are based on the rules of the physical world rather than patterns in the text. They can perform quantum chemistry calculations and simulate molecular dynamics and microkinetics, which is the study of how chemical reactions unfold at the molecular level. This is important because it tells researchers how candidate molecules are likely to behave before anyone sets foot in a lab.

“Trained on lab data and real-world scientific equations, LQMs are AI models designed for the quantum economy, a $50 trillion sector that includes biopharma, financial services, energy and advanced materials,” the company said in a press release.

Chai Discovery and Isomorphic Labs—both well-funded bets on better models—have focused on science. SandboxAQ is focused on those who can actually use it.

“For the first time, we have limits [quantitative] “A model on a parametric LLM can be accessed by anyone in natural language,” Nadia Harhin, general manager of AI simulation at SandboxAQ, told TechCrunch. Previously, users of LLMs in SandboxAQ had to provide their own digital infrastructure to run the models.

SandboxAQ clients tend to be computational scientists, research scientists, or experimentalists. Generally, these people work in large pharmaceutical or industrial companies and are looking for new materials that can become marketable products.

“Our clients come to us because they’ve tried all the other software out there, and the complexity of their problem was such that it didn’t work or didn’t yield positive results for them when that real-world translation started,” Harhin said.