---
title: "AI Inference Gateway"
url: "/docs/ai-inference-gateway/"
kind: "hub"
summary: "Route inference requests on Kubernetes with Gateway API, InferencePool, and an endpoint picker that selects from ready model server Pods."
---

> Arion site index: [/llms.txt](/llms.txt)

# AI Inference Gateway

Route inference requests on Kubernetes with Gateway API, InferencePool, and an endpoint picker that selects from ready model server Pods.

Use Kubernetes Gateway API and the Inference Extension to route to self-hosted model servers. The Arion controller translates an InferencePool into a pool of ready Pods; an external endpoint picker chooses the destination for each request through `ext_proc`.