Anemoi: A European framework for operational AI in weather and climate

Summary

Artificial intelligence (AI) is rapidly transforming weather and climate science across Europe. What is changing is not just how models are built, but how information is produced, delivered and used by national meteorological services, public authorities and downstream users. We are shaping this transformation with our Member and Co‑operating States, building on decades of shared investment in data, infrastructure and operational expertise. Anemoi plays a key role in this transformation. Developed by ECMWF together with several national meteorological services across Europe, Anemoi is a joint European framework enabling the development, training and operation of large‑scale AI models for weather and climate. Rather than focusing on individual models, Anemoi provides shared foundations – datasets, workflows and software infrastructure – that allow AI to move from research into pre‑operational and operational use across national meteorological services and at ECMWF. From data to decisions: why Anemoi matters Anemoi supports rapidly evolving machine‑learning (ML) techniques while maintaining the reliability, traceability and robustness required for operational services. Achieving this at scale requires much more than neural network architectures alone. It requires: - AI‑ready datasets built with deep Earth‑system domain knowledge - Robust, reusable workflows that run efficiently on distributed European HPC systems - Integrated tooling for training, inference, verification and monitoring Anemoi was created to address this shared need. Its focus on end‑to‑end workflows allows new AI capabilities to be developed, tested and deployed collaboratively, without each user having to rebuild the full software stack. This supports innovation while preserving scientific rigour and operational consistency across Europe. What is Anemoi? Anemoi is an end‑to‑end software framework for AI weather and climate applications. It covers the full lifecycle from data preparation to model training and large‑scale inference. The framework is modular, Python‑based and designed to run efficiently on modern high-performance computing (HPC) architectures, including EuroHPC systems. The framework brings together three tightly integrated components. 1. Anemoi datasets Anemoi produces and manages AI‑ready datasets optimised for ML workloads. These datasets differ fundamentally from traditiona…