Skip to content

GraphCast

A deep learning model for global medium range weather

Visit official site

What graphcast is

The model is a Google DeepMind deep learning system for global medium range weather forecasting. Its repository covers running and training GraphCast and GenCast.

What you can do with it

  • Load data or random weights for a run
  • Generate predictions from pretrained snapshots
  • Compute loss and gradients

Getting started

  • Open the GitHub page and read the docs
  • Download data sets such as ERA5
  • Set up Google Cloud and open a Colab notebook

Pros & cons

✓ What we like

  • Google Cloud setup guidance is included
  • Multiple pretrained models fit different resolutions

! What to watch out for

  • ERA5 data must be downloaded separately
  • Cloud setup such as a TPU VM is part of the workflow

FAQ

Who developed it?

Google DeepMind, as stated in the material.

What sample inputs are provided?

The repository includes pretrained weights, normalization statistics and example input data.

Last reviewed: 2026-11-15

More LLM API platform tools

View all →

How we review