Buoyant open-sources Kinference inference control plane for Kubernetes and vLLM

Buoyant has open-sourced Kinference, an Apache 2.0 inference control plane for vLLM on Kubernetes.

The news was first reported by MarkTechPost on November 7, 2026. According to the article, Kinference adds declarative routing policies, KV-cache-aware load balancing, and per-tenant metrics on top of vLLM workloads running on Kubernetes. Buoyant also claimed that its own benchmark measured a 95% reduction in median TTFT, though no third-party verification was provided.

At this point, Kinference exists only in the form of this single MeltwaterPost announcement: no repository details, release artifacts, contributor list, or follow-up coverage were included in the available material. That means the project's actual code availability and community adoption remain unconfirmed, and the TTFT figure stands solely as a vendor-reported benchmark rather than an independently validated result.

The announcement was framed as available under the permissive Apache 2.0 license, and the piece did not state any paid tier or commercial restrictions, so the scope within which developers may adopt the project is not yet fully documented. Further developments will be reported here once they are verified in public sources.