HP Z Runtime0.51.2

Run zrt commands to prepare a vLLM environment and serve large language models from a Linux server. The command-line interface handles model lifecycle tasks, from obtaining a model to running and observing its workload. It exposes an OpenAI-compatible HTTP API, allowing clients and services that use that interface to send inference requests.

Models can be pulled from Hugging Face, Azure ML, or MLflow, then served, used for interactive chat, and monitored through the same CLI. Automatic GPU memory management helps coordinate model serving on hosts running GPU workloads. By combining these functions with the vLLM inference engine, zrt provides a production-oriented workflow for developers and data science teams operating language-model services.

HP Snap Maintainers publishes HP Z Runtime for Linux as commercial Developer Tools software. It is intended for Linux servers with GPU workloads, particularly when a team needs a controlled command-line path from environment preparation through ongoing model operation.

The host must provide Linux and snapd, and supported architectures are amd64 and arm64. The current version is 0.51.2.