Kinference Unveiled for Inference Control in Kubernetes

SAN FRANCISCO, Nov. 4, 2026 /CourierPR/ -- Buoyant, the creator of Linkerd, today announced the general availability of Kinference, an open-source inference control plane for Kubernetes. Kinference, which turns open-weight models and GPUs into a self-hosted inference platform, enables organizations to deliver safe, reliable, and cost-effective inference across multiple models, GPUs, and machines, while providing the governance and policies necessary for enterprise use at scale.

Kinference, built on top of the open-source vLLM inference engine, adds a unified layer of routing, load balancing, and policy-driven control to the existing foundation of Kubernetes. This new layer allows organizations to provide a true, internal "inference platform" capable of managing inference guardrails and controls, lowering the cost of inference, and enabling intent-based routing. Additionally, Kinference supports per-tenant usage and cost attribution, providing organizations with a full picture of their inference spend.

Kinference's features include:

Kinference has already been tested in public benchmarks, where it was able to cut median time-to-first-token (TTFT) by 95% and median end-to-end latency by 45% of requests on one workload type. This performance improvement underscores Kinference's potential to significantly enhance the efficiency and reliability of in-house inference operations.

As adoption of generative AI continues to grow, organizations are increasingly moving inference in-house, often with open-weight models and GPUs under their control. Kinference's ability to provide a unified layer of control and governance for these in-house operations makes it a valuable tool for enterprises looking to manage and optimize their AI infrastructure.

Buoyant, the creator of Linkerd, is excited to launch Kinference as a reliable, high-performance, and secure option for platform engineers. Kinference is lightweight, easy to deploy, and can be run without a team of specialists, making it accessible to a wide range of organizations. Kinference is agnostic to LLM models, cloud providers, and GPU types, allowing organizations to use it with their existing infrastructure and models.

To see Kinference in action, join Buoyant for a livestream on Thursday, November 19, 2026, at 9 AM PT. Register here: [insert link]