NVIDIA announces Blackwell Ultra and Dynamo inference software

NVIDIA announced Blackwell Ultra, including the GB300 NVL72 and HGX B300 NVL16 systems, alongside NVIDIA Dynamo open-source inference software. The company framed the platform around training and test-time inference for reasoning, agentic, and physical-AI workloads. Its release also described Spectrum-X networking updates and reported generation-over-generation performance comparisons for selected large-language-model inference workloads.

Original source date: . Hypler briefing published October 6, 2026.

Topics: NVIDIA, Blackwell, inference, reasoning

NVIDIA logo

Generated editorial illustration. Inference scheduling illustrated with a generic liquid-cooled accelerator rack; not a depiction of actual Blackwell hardware.
Generated editorial illustration. Inference scheduling illustrated with a generic liquid-cooled accelerator rack; not a depiction of actual Blackwell hardware. Credit: Hypler / AI-generated editorial illustration. Hypler editorial asset; not vendor or event photography
Generated illustration of connected accelerator trays and cooling channels; not a vendor facility.
Generated illustration of connected accelerator trays and cooling channels; not a vendor facility. Credit: Hypler / AI-generated editorial illustration. Hypler editorial asset; not vendor or event photography

Engineering relevance

Reasoning workloads make infrastructure choices more visible because latency, memory, networking, and queue behavior affect the user-facing system. Vendor platform announcements can guide capacity research, but benchmark figures are not deployment guarantees. System records should capture the model, hardware, serving stack, evaluation method, and workload constraints that produced any operational conclusion.

Platform announcement

NVIDIA introduced Blackwell Ultra as a successor platform for AI factories, naming the GB300 NVL72 rack-scale system and HGX B300 NVL16. The release connected the hardware to training and test-time scaling inference, where a system applies additional computation during inference to improve a response.

Dynamo software

The company also announced NVIDIA Dynamo, an open-source inference framework intended to scale reasoning services. NVIDIA described it as a way to improve throughput and response time while reducing serving costs. The announcement paired it with networking updates, not as a standalone claim about end-to-end application performance.

Original source

This is a historical source briefing, not a statement of current availability or Hypler deployment.