• Picks
  • Discover
  • Alternatives
  • About

© 2026 Willet

  • Newsletter
  • About
  • FAQ
  • Terms
  • Privacy & Cookies
Join the Discord (opens in a new tab)
Google DeepMind avatar

WeatherNext

Google DeepMind's AI models for global weather and cyclone forecasting

  • AI models and runtimes
  • Education
View on GitHub

What it is

Code and pretrained weights for WeatherNext 2, a medium-range global atmospheric forecasting model that also tracks tropical cyclones, along with the earlier GraphCast and GenCast models. It loads initial conditions such as HRES data, runs autoregressive rollouts at 0.25° or 1° resolution, and includes a cyclone tracker plus training loss code. Weather researchers and forecasters use it to run, evaluate, or fine-tune the published models on TPUs or GPUs, starting from a Colab notebook.

Images

Facts

Stars
7,718 GitHub stars.
7,718

Forks
997
Commits
55
Contributors
13
Language
Python
License
Apache-2.0
Last commit
Sep 4, 2026
On GitHub since
July 2023
First release
October 2023
Platforms
library
Self-hostable
No
Alternative to
ECMWF HRES
Tags
  • research-tools
  • open-weights
  • machine-learning
  • time-series
  • scientific-computing

Last synced from GitHub on Oct 20, 2026

The 10 best overlooked open-source projects, every Monday.

Share and more

Comments

No comments yet. Start the conversation.

What people are saying

Updated Oct 20, 2026
  • Bluesky“the code is open source. I checked it myself. it uses Jax and Haiku libraries for its machine learning capabilities. it does not hook into LLMs which are for…”Aug 20, 2026 · 5 likes · 1 reply (opens Bluesky in a new tab)
  • Bluesky“Guide: Repo: TPU Developer Hub:”Sep 8, 2026 · 1 like · 0 replies (opens Bluesky in a new tab)