A framework for general-purpose AI model categorisation : collection of external scientific studies on general-purpose AI models under the EU AI Act

This report proposes a framework for categorising AI models as General-Purpose AI (GPAI) models as defined in the EU AI Act, based on their capabilities and generality. It breaks down the core components of the GPAI model definition into measurable elements, focusing on four primary cognitive domains: (1) Attention and Search, (2) Comprehension and Compositional Expression, (3) Conceptualisation, Learning and Abstraction, and (4) Quantitative and Logical Reasoning. The report suggests using the Annotated Demand Levels (ADeLe) procedure to evaluate AI models' capabilities in these domains, and provides a methodology for combining domain-level scores into a single measure of generality. The framework is illustrated with empirical results from existing models, and policy recommendations are made for selecting thresholds and metrics for GPAI model categorisation.