Meet the New Standard for High-Performance, Low-Cost Inference: NVIDIA Dynamo 1.0 is now available to DigitalOcean Customers

NVIDIA Dynamo 1.0, announced at GTC, is now available to DigitalOcean customers. It delivers up to 7x inference performance improvement on NVIDIA GB200 NVL systems through key features: KV-aware routing, disaggregated prefill/decode serving, and memory offloading via a KV Block Manager. Paired with DigitalOcean's Agentic Inference Cloud and Managed Kubernetes, Workato achieved 67% higher GPU throughput, 79% lower latency, and 67% lower model cost using half the GPUs. Customers can deploy Dynamo 1.0 as a container on Droplets or via DigitalOcean Kubernetes with inference runtimes like vLLM, SGLang, or TensorRT-LLM.