This repository serves as a curated collection of Large Weather Models (LWMs), aiming to centralize information on papers, datasets, and code for researchers and practitioners in AI for Earth and AI for Science. It provides a structured overview of the rapidly evolving field of data-driven weather forecasting.

How It Works

The project acts as a knowledge hub, aggregating links to research papers, benchmark datasets, and open-source code repositories for various LWMs. It categorizes models by their development origin, publication status, and licensing, offering a comparative view of different approaches in AI-driven weather prediction.

Quick Start & Requirements

This is a curated list, not a runnable project. Users are directed to individual model repositories for installation and execution. Links to papers, code, and datasets are provided within the README.

Highlighted Details

• Comprehensive list of LWMs including MetNet, FourCastNet, Pangu-Weather, GraphCast, ClimaX, and FengWu.
• Detailed table comparing models by developer, release date, publication, licensing, and availability of weights.
• Extensive news section tracking recent advancements and operational deployments of LWMs.
• Links to relevant datasets like WeatherBench, ERA5, and SEVIR.

Maintenance & Community

The repository welcomes contributions. It lists major research institutions and companies involved in LWM development, such as Google DeepMind, NVIDIA, Microsoft, Huawei, and ECMWF.

Licensing & Compatibility

Licenses vary significantly across listed models, ranging from permissive MIT and Apache 2.0 to more restrictive non-commercial licenses (e.g., CC-BY-NC-SA 4.0) for model weights. Users must consult individual model licenses for usage rights.

Limitations & Caveats

This is an informational resource, not a unified framework. Users must navigate to individual model repositories for code execution, which may have diverse dependencies and setup requirements. Licensing for model weights is often not specified or is restrictive.