“performance optimization”

Mistral Small 3 is a 24B-parameter language model optimized for low-latency performance across common AI tasks.