EU Frontier AI Initiative: Advanced Scaling Frameworks for High-Performance AI Models (RIA) - Apply AI

What is this call about?

The topic funds one research project on methods and tools for predictable scaling of AI models, helping European AI labs use computing resources efficiently.

What does this call fund?

Scope

The topic funds research on the theoretical basis, methods and tools for predictable scaling of AI models, for both training and inference. The project should deliver a toolset that supports scaling of AI models and complements training-focused efforts.

Possible research areas

• automated derivation of scaling laws across modalities and tasks
• automated search for model architectures, including sparsity, memory efficiency and long-context handling
• optimisation of dataset mixtures and automated data quality assessment
• stable training at large scale and automated hyperparameter tuning

Further expectations

Methods should work across different hardware architectures. The project should support advanced AI labs and high-performance computing centres, link to AI Factories and Data Labs, share results through the AI-on-demand platform and cooperate with the Partnership on AI, data and robotics and the GenAI4EU hub. It forms part of the Frontier AI Initiative.

Expected outcomes

Reliable prediction of model performance before large compute investment, more competitive European AI models and scientifically grounded scaling methods.

Technology readiness

Activities are expected to reach TRL 4 to 5 by the end of the project.

Further expectations

• Proposals should describe the link between their activities and topic HORIZON-CL4-2027-04-DIGITAL-EMERGING-11.

When is the deadline?

16 February 2027 (103 days)