NVIDIA has announced NVIDIA Dynamo, a new open-source framework for low-latency distributed inference aimed at scaling AI models efficiently. This release was made during GTC 2025, promising high throughput and enhanced performance for AI applications.