Lightweight F1 variant for resource-constrained deployments.
Blazing-fast inference for generative AI