Hugging Face Ships One-Command vLLM Servers on HF Jobs — OpenAI-Compatible Endpoints With No Infra Provisioning

Summary

Hugging Face introduced a streamlined path to stand up a private, OpenAI-compatible vLLM inference server using its Jobs infrastructure with a single `hf jobs run` command — specifying a hardware flavor, port exposure, and a timeout. Requests route through HF's public jobs proxy, authentication is via HF tokens, and the endpoint can be queried with curl or the standard OpenAI Python client. Billing is pay-per-second, charging only while the job is actually running. The post (by Quentin Gallouéde