AI that doesn't guess.
It computes.
Large Quantitative Models learn from physics, chemistry and measurement — not the internet. They screen molecules, predict battery life and simulate engineering designs in seconds. This is the independent guide to the field.
One term. Three ideas.
Language models made AI fluent. Quantitative models make it exact. Here is the category in one screen.
A large-scale AI system trained on numerical, physical and scientific data — often generated from equations — whose predictions can be checked against measurable reality.
- 01Grounded in physicsTrained on first-principles simulations and lab measurements, not web text.
- 02Built for a targetOptimises toward a property: binding affinity, cycle life, drag, risk.
- 03Hybrid by designModular stacks where an LLM is the interface and the LQM does the computing.
Large Language Model LLM
Large Quantitative Model LQM
Error figures from SandboxAQ × NOVONIX (2024), using two cycles of high-precision data. Curves are illustrative. Dashed line = measured truth.
The numbers behind the hype.
Every figure below comes from a public source and links to the article where we unpack it.
Less time to predict battery end-of-life
SandboxAQ × NOVONIXAccuracy gain over capacity-based models
SandboxAQ × NOVONIXProtein–ligand structures in the open SAIR dataset
SandboxAQ, 2025PhysicsX valuation after its June 2026 round
PhysicsXWho is building it.
The field is small, well funded and moving fast. These are the names that define it today.
SandboxAQ
The company most associated with the term. Scientific LQMs for drug discovery, materials and batteries, now distributed through Google Cloud Marketplace and reachable from AI assistants.
PhysicsX
AI-native engineering platform for aerospace, semiconductors, automotive and energy. Developing Large Physics Models; backed by Temasek, NVIDIA, Siemens and Applied Materials.
Siemens Simcenter
PhysicsAI brings geometric deep learning and reduced-order models into established engineering simulation workflows — the incumbent path into quantitative AI.
Aviso
Uses LQMs as the reasoning layer for revenue forecasting and pipeline risk, with LLMs extracting signals from emails and meetings — the hybrid pattern in everyday software.
Long reads for serious readers.
Sourced, specific and written for people who build, buy or fund quantitative AI.
Why Gartner put LQMs first for 2027
On October 20, the most-quoted trend list in enterprise IT opened with a term most CIOs had never heard. Here is what the call means, and what it doesn't.
Read article →What is a Large Quantitative Model?
A field guide to the AI category that Gartner just named the number-one strategic technology trend for 2027 — and why it exists at all.
Read article →LQMs in the wild: five deployments
Batteries, catalysts, drug discovery, industrial engineering and revenue forecasting — what quantitative models are already doing outside the slide deck.
Read article →Words vs. numbers: LLM and LQM
They are not rivals. They are two halves of the same stack — and understanding the seam between them is the most valuable architectural skill of the next five years.
Read article →The road to 2030
A detailed outlook for Large Quantitative Models: where the money is going, which industries move first, what the stack will look like, and the risks that could slow it all down.
Read article →How we got here.
From an Alphabet side project to the top of the most-quoted trend list in enterprise IT.
SandboxAQ spins out of Alphabet
A company built at the intersection of AI and quantum-inspired computing starts to popularise the term "Large Quantitative Model".
Battery life predicted in days
With NOVONIX data, LQMs cut end-of-life prediction time by 95% — an early, measurable proof point.
$450M Series E
SandboxAQ raises at a reported $5.6B valuation, signalling investor conviction in quantitative AI.
Capital and distribution
PhysicsX raises $300M; SandboxAQ announces LQMs on Google Cloud Marketplace, starting with AQCat.
Gartner ranks LQMs #1 for 2027
The category gets its institutional moment — and a place in thousands of board decks.
Three horizons to 2030.
Our analysis of how the category is likely to mature. The full reasoning is in The road to 2030.
Naming & sorting
- Vendors relabel; buyers need filters
- Pharma, batteries, chemicals budget real programmes
- First quantitative benchmarks appear
The platform layer
- Composable model zoos per domain
- Validation & uncertainty become products
- Proprietary data decides the winners
Closed loops
- Models propose, labs test, data returns
- Simulation becomes a live design tool
- Convergence with world models
Building quantitative AI? Let's talk.
Join a private network of people working on Large Quantitative Models. Share research, publish with us, and meet teams that are hiring.
Join the network →