NVIDIA's open and NVFP4-optimized models for reasoning, agentic, and long-context workloads. Fast and memory-efficient with native Blackwell acceleration. A strong fit for serving frontier models at scale, cost-effectively.