Pre-Publication V2. An artificial intelligence model trained on large-scale, domain-diverse datasets and intended to perform a wide range of tasks across subject areas and modalities (including text, image, audio, video, and code without being designed for any single specific application). Unlike task-specific models, GPAI models aim to exhibit broad capability generality and may demonstrate emergent abilities not explicitly present in their training objectives. They can function as standalone systems or serve as foundation components for downstream applications, products, and services.
Deliberation Summary:
Edits made for clarity and consistency
Pre-Publication V1. An artificial intelligence model trained on large-scale, domain-diverse datasets to perform a wide range of tasks across subject areas and modalities including text, image, audio, video, and code without being designed for any single specific application. Unlike task-specific models, GPAI models exhibit broad capability generality and may demonstrate emergent abilities not explicitly present in their training objectives. They can function as standalone systems or serve as foundation components for downstream applications, products, and services.