The world's top weather forecasting model just went open source
What does it take to predict the weather five days out? At the European Centre for Medium-Range Weather Forecasts (ECMWF), the answer is a numerical model called the Integrated Forecasting System (IFS) -- one of the most accurate global weather prediction systems ever built, developed over nearly four decades in partnership with national meteorological services across Europe.
Until now, a portable version of that model, called OpenIFS, was available only under institutional licenses. Researchers at universities could use it, but the process involved paperwork, restrictions, and a version that lagged behind the operational system. As of March 5, 2026, that has changed. ECMWF has released OpenIFS as fully open-source software on GitHub.
Why open-source matters for weather models
The shift eliminates bureaucratic barriers to access. Anyone -- university researchers, private-sector meteorologists, educators, citizen scientists -- can now download, modify, and run the same fundamental forecasting system that produces ECMWF's operational predictions. The code can be cited in publications, shared freely as part of open-access journal papers, and even run on a laptop for educational purposes.
"Opening the model will broaden access, deepen collaboration and accelerate innovation across our community," said ECMWF Director-General Florian Pappenberger. "The Integrated Forecasting System is a shared achievement of our Member and Co-operating States, built on scientific excellence, operational expertise and sustained long-term investment."
The model's roots trace back nearly 40 years to a collaboration with Meteo-France, and the IFS/Arpege code has been updated through more than 50 versions. Virginie Schwarz, President and CEO of Meteo-France, described the release as sharing "the most powerful code in the world" with the international community.
Closing the lag between research and operations
One practical benefit is currency. The operational IFS updates annually, but previous OpenIFS releases were less frequent, creating a gap between what researchers could access and what forecasters actually used. Open-sourcing allows ECMWF to keep OpenIFS much closer to the current operational version.
"Just as importantly, [going open source] will help OpenIFS users stay current," said Mike Sleigh, Head of Integrated Forecast Systems. "IFS updates annually, but OpenIFS releases have until now been less frequent. Going open source will help us to keep OpenIFS much closer to the operational version of IFS, providing up-to-date code and avoiding the longstanding lag between research and operations."
The change also supports reproducible science. When researchers can share the exact model code they used alongside their papers, other scientists can replicate their results -- a persistent challenge in computational weather and climate research where proprietary or version-specific models make reproducibility difficult.
Fifteen years of community building
OpenIFS is not new. The project has operated for 15 years, building a growing community of users who have applied the model to atmospheric research, climate modeling, numerical weather prediction education, and more. The ECEarth-4 climate model, development of single-precision IFS, and research into idealized atmospheric modeling have all drawn on OpenIFS.
"OpenIFS has become a popular teaching tool for numerical weather prediction and modelling at universities across Europe," said Marcus Koehler, an ECMWF scientist on the project. "But the change to open source really provides true open access to anyone, for any purpose, including the potential for industry collaboration."
What this does and does not include
OpenIFS is a portable version of the IFS forecast model. It allows users to run atmospheric simulations and generate forecasts. It does not include ECMWF's full operational infrastructure -- the data assimilation systems, observational data streams, and high-performance computing environment that produce the official forecasts distributed to meteorological services worldwide.
Running a competitive weather forecast still requires massive computational resources and access to global observational data. What OpenIFS provides is the core scientific engine -- the physics parameterizations, dynamical core, and numerical methods -- that can be used for research, education, model development, and comparison studies.
The release follows the open-source model already adopted by other global weather centers and reflects a broader trend in computational science toward transparency and accessibility. For a model built on decades of multinational investment, making it freely available represents a philosophical statement as much as a technical one: the best tools for understanding the atmosphere should not be locked behind institutional gates.