GraphCast / WeatherNext
Google DeepMind's machine-learning weather models that forecast the globe in minutes — now adapted into NOAA's operational systems.
Overview
GraphCast is a graph-neural-network model from Google DeepMind that produces accurate medium-range (10-day) global weather forecasts in minutes rather than the hours traditional physics-based supercomputing needs. Its code and trained weights are openly published. WeatherNext is the productised model family derived from that research, offered via Google Cloud and Earth Engine. In December 2025, NOAA deployed operational AI forecast systems (AIGFS/AIGEFS) fine-tuned from GraphCast on its own data.
Pricing
Pricing shown for reference only. These figures reflect RECATOOLS research as of 24 Jul 2026 and may be out of date or incomplete. This is not financial or purchasing advice — always confirm the current price on the provider’s official website before making any decision.
ASEAN Perspective
GraphCast / WeatherNext in Southeast Asia
ASEAN-region availability and pricing notes coming soon. Drop the editorial team a note via /contact/ if you can supply local context (Singapore/Malaysia/Indonesia/Thailand/Vietnam).
What this is for: Fast, accurate ML-based global weather forecasting for research and operational meteorology.
Who this is for: Meteorologists, climate researchers, and developers building forecasting products.
Availability: GraphCast is open-source (weights on GitHub); WeatherNext data is available through Google Cloud and Earth Engine, and a GraphCast-derived model is now operational at NOAA.
What people say
GraphCast earned serious scientific credibility fast: its 2023 Science paper showed it beating the ECMWF high-resolution deterministic forecast on a large majority of variables while running in minutes on a single machine. The clearest vote of confidence is operational adoption — NOAA's National Hurricane Center began referencing GraphCast-class AI guidance in its forecast discussions, and in late 2025 NOAA deployed GraphCast-derived operational systems (AIGFS/AIGEFS). Independent evaluations over regions like Brazil have broadly reproduced its medium-range skill.
Meteorologists raise substantive limitations. GraphCast is a "black box" that predicts accurately without explaining mechanisms, which matters when forecasters need to reason about wind shear or convective dynamics. Trained on 1979-2018 reanalysis, it can under-represent unprecedented, climate-driven extremes, and studies note it tends to blur or under-forecast extreme precipitation and small-scale structure. Recent work also finds fundamental problems using ML models in data assimilation — their adjoint/tangent-linear behaviour can produce unrealistic corrections.
Crucially, ECMWF, NOAA and the Met Office all frame these models as augmenting, not replacing, physics-based systems and human forecasters. The reception is genuine enthusiasm tempered by the understanding that speed and skill do not yet equal interpretability or reliability on the tail events that matter most.
Summary of public user & expert reviews, compiled by RECATOOLS.
About this listing
This entry was compiled from publicly available data including GraphCast / WeatherNext's official website, press releases, documentation, and reputable third-party publications. RECATOOLS is not affiliated with GraphCast / WeatherNext unless explicitly stated.
Third-party AI tools update their pricing, features, availability, and policies frequently. Information here may be outdated by the time you read this — we make reasonable efforts to keep listings current, but cannot guarantee absolute accuracy.
For the latest details, please refer to GraphCast / WeatherNext directly →
Spotted something out of date? Suggest an update →
More in Research & Data
