歐盟與EuroHPC推出前沿AI大挑戰

• Models must achieve computational capacity equivalent to at least 400 billion parameters[1] • Efficient, modular architectures such as Mixture-of-Experts (MoE) are expected to set new benchmarks for performance and efficiency[1] • Selected projects will utilize EuroHPC computing resources, specifically AI-optimized exascale or pre-exascale supercomputers[2] • The models are designed as general-purpose systems capable of adapting across domains with minimal modification[1] • Strategic computing access allocation: up to 2.5% of overall EuroHPC computing resources for one year[1][2]

The Frontier AI Grand Challenge represents a critical EU strategic initiative to reduce Europe's dependence on non-European AI systems and establish technological sovereignty in frontier AI development. By combining world-class EuroHPC infrastructure with European AI talent, the initiative aims to reinforce the EU's competitive position in global AI competition and foster a trustworthy, human-centric AI ecosystem[1]. The emphasis on open models and broad accessibility across public, scientific, and business sectors is designed to accelerate innovation across key European industries while strengthening the startup and scale-up ecosystem. This complements the broader AI Continent Action Plan and positions Europe to compete with leading global AI powers.