Google Cloud partners with SandboxAQ to bring specialist AI models built for scientific research

Google Cloud expands marketplace with specialist AI models for science

Google Cloud is partnering with SandboxAQ to offer large quantitative models through its marketplace, marking a strategic shift in how AI models serve scientific research 1

1

. Unlike traditional large language models trained on text, these specialist AI models for science are built on numerical data, scientific equations, and real-world laboratory measurements 2

2

. The move acknowledges a fundamental limitation: while language models excel at generating fluent text, they prove surprisingly unreliable when precision with numbers matters most.

SandboxAQ introduces AQCat and AQPotency LQMs for research teams

SandboxAQ, which began life at Alphabet in 2016 before spinning out in 2022, will make two LQMs available through Google Cloud Marketplace 2

2

. The AQCat model, launching in Q3 2025, targets catalyst discovery and adsorption energy calculation, enabling researchers to identify promising candidates for materials science before committing costly resources to full evaluation. SandboxAQ notes that catalysts underpin more than 90% of all commercially produced chemical products and directly impact initiatives including green hydrogen, sustainable aviation fuel, fertilizer production, and plastics recycling 2

2

. The AQPotency model, arriving later in 2026, focuses on pharmaceutical R&D by computationally identifying binding molecules for drug discovery, evaluating thousands of candidates in a fraction of traditional screening time and cost 2

2

.

Gemini for Science combines language models with quantitative precision

Researchers using Google Cloud will be able to combine these quantitative models with Gemini, leveraging the language model for reasoning and interface while the quantitative model handles underlying science 1

1

. Google paired the marketplace expansion with Gemini for Science, a bundle of tools including its AI co-scientist, AlphaEvolve coding agent, empirical research assistant, and NotebookLM, designed to accelerate routine steps in scientific research workflows rather than replace scientists 1

1

. Jack D Hidary, CEO of SandboxAQ, stated that pairing Gemini's reasoning with the quantitative precision of LQMs creates a powerful combination that puts first-principles science directly into researchers' hands 2

2

.

Why this matters for enterprise AI infrastructure and scientific discovery

The partnership reflects intensifying competition among hyperscalers to capture high-value segments that general chatbots cannot serve effectively. By offering specialist models through its marketplace, Google Cloud can address demand for scientific research and industrial R&D without building every domain model itself 1

1

. This approach fits Google's broader pattern of applying AI to laboratory results, following DeepMind's protein-structure work that reshaped drug development and materials discovery efforts that found more new materials in a year than science had catalogued in its entire history 1

1

. Brian Goldstein, vice-president of strategic AI and independent software vendors at Google Cloud, emphasized that bringing these models to the marketplace helps healthcare researchers accelerate drug discovery and address critical gaps 2

2

. The capabilities are already in use by partners in private preview for real-world R&D, though specific organizations and results remain undisclosed 1

1

. What was once confined to specialist labs now becomes something research teams can rent, with the private previews determining whether this produces genuine discoveries or simply faster processes. SandboxAQ, chaired by former Google CEO Eric Schmidt, claims its models can deliver critical advances across life sciences, financial services, and navigation, and is already integrated with Anthropic's Claude 2

2

. As DeepMind's drug-discovery spinoff Isomorphic Labs moves toward trials and rivals race to convert algorithmic promise into treatments that work outside benchmarks, this partnership positions Google as the AI infrastructure underneath that race, particularly for semiconductor manufacturing and other industries where numerical accuracy determines success 1

1

.