01LQM platforms
SandboxAQ
The company that uses the term Large Quantitative Models for AI trained on physics, chemistry and proprietary scientific data.
Best for: The main commercial LQM provider.
Visit SandboxAQ →Large Quantitative Models (LQMs)
AI models grounded in numbers, physics and scientific data rather than text: SandboxAQ's LQMs, time-series foundation models, scientific foundation models and finance-focused models.
16models
4Categories
5FAQs answered
02LQM platforms
SandboxAQ drug discovery
SandboxAQ's LQM-based platform that predicts how molecules behave before they are synthesised in a lab.
Best for: Computational chemistry for drug discovery.
Visit SandboxAQ drug discovery →03LQM platforms
SandboxAQ material discovery
SandboxAQ's LQM-based approach to simulating and screening candidate materials.
Best for: Materials and chemistry research.
Visit SandboxAQ material discovery →04Time-series foundation models
TimesFM
Google Research's pretrained time-series foundation model for zero-shot forecasting.
Best for: Forecasting without training per series.
Visit TimesFM →05Time-series foundation models
Chronos
Amazon's family of pretrained time-series forecasting models that treat values as tokens.
Best for: Probabilistic forecasts from pretrained models.
Visit Chronos →06Time-series foundation models
Moirai (uni2ts)
Salesforce AI Research's universal time-series forecasting models and the uni2ts library.
Best for: Forecasting across frequencies and domains.
Visit Moirai (uni2ts) →07Time-series foundation models
IBM Granite time series
IBM's compact open time-series foundation models and tooling for forecasting and anomaly detection.
Best for: Small time-series models for production.
Visit IBM Granite time series →08Time-series foundation models
Nixtla TimeGPT
Nixtla's time-series foundation model, offered as an API alongside open-source forecasting libraries.
Best for: Hosted forecasting with minimal setup.
Visit Nixtla TimeGPT →09Time-series foundation models
Kronos
Open foundation model for the language of financial markets, trained on candlestick (K-line) data.
Best for: Research on market-data foundation models.
Visit Kronos →10Scientific foundation models
AlphaFold 3
Google DeepMind's model for predicting the structures of proteins and their interactions with other molecules.
Best for: Biomolecular structure prediction.
Visit AlphaFold 3 →11Scientific foundation models
MatterGen
Microsoft's generative model for designing new inorganic materials.
Best for: Generating candidate materials.
Visit MatterGen →12Scientific foundation models
Aurora
Microsoft's foundation model for the Earth system, used for weather and atmosphere forecasting.
Best for: Weather and climate prediction.
Visit Aurora →13Scientific foundation models
BioEmu
Microsoft's model for generating the range of shapes a protein can take.
Best for: Protein dynamics.
Visit BioEmu →14Scientific foundation models
Evo 2
Arc Institute's open genomic foundation model trained on DNA sequences from many organisms.
Best for: Genomics research.
Visit Evo 2 →15Scientific foundation models
FAIR Chemistry (fairchem)
Meta FAIR's open-source library and models for machine-learning force fields and chemistry.
Best for: Fast molecular and materials simulation.
Visit FAIR Chemistry (fairchem) →16Finance models
FinGPT
Open-source project for finance-focused language models and data pipelines.
Best for: Financial text and sentiment research.
Visit FinGPT →Choosing the right approach
Match your requirement to a starting point, then compare vendors.
| Chemistry, drug or materials discovery | SandboxAQ's LQM platforms or scientific models such as MatterGen and fairchem |
|---|---|
| Forecast many numeric series | A time-series foundation model such as TimesFM, Chronos or Moirai |
| Hosted forecasting | An API service such as Nixtla TimeGPT |
| Protein, genome or structure work | AlphaFold 3, BioEmu or Evo 2 |
| Weather and climate | Aurora |
| Market data research | Kronos, with strong validation and risk controls |
Related pages on PeterIndia.net
Neighbouring directories and hubs.
Time-Series Forecasting AlgorithmsClassical and modern forecasting methods.AI Model ArchitecturesGPT, MoE, LRM, SLM and LAM explained.Quantum ComputingRelated quantum technologies.Large Language ModelsThe directory of LLMs.Deep LearningNeural networks behind these models.Machine LearningThe machine learning hub.Python Libraries for Time-Series AnalysisLibraries for numeric series.Generative AIThe generative AI hub.
Frequently asked questions
Quick answers.
What is a large quantitative model?
The term was popularised by SandboxAQ for AI models trained on physics, chemistry and quantitative scientific data so they can predict real-world behaviour of molecules and materials, rather than model language.
How is an LQM different from an LLM?
An LLM learns patterns in text. An LQM is built from equations, simulation and measured data in a scientific domain, and can be combined with LLMs as an interface.
Is LQM an established category?
Not yet. It is a vendor-led term and is used loosely. This page also lists related quantitative foundation models, such as time-series and scientific models, that share the idea.
Are time-series foundation models LQMs?
Not strictly. They are numeric foundation models trained on large collections of series. They are included because they apply the same idea of large pretrained models to quantitative data.
Can these models be trusted for decisions?
Treat outputs as predictions to be validated. Scientific models need lab or simulation checks, and forecasting and market models need back-testing and risk limits.
Descriptions are summaries of each vendor's own positioning, checked against its website on 4 October 2026. Product names and offerings change often, so confirm details on the vendor site before deciding.