What is LLM gateway?
An LLM gateway is a single API endpoint that routes requests to many language models behind one key, one bill, and one interface.
An LLM gateway sits between your application and the underlying models. Instead of integrating each model separately — its own SDK, key, billing, and quirks — you call one endpoint and select a model with a string. The gateway handles authentication, request translation, usage metering, and (in the better ones) automatic failover when a model is unavailable.
Gateways matter because the model landscape changes weekly. New models ship, prices move, and the best model for a task shifts. A gateway lets you switch models with a one-line change instead of a re-integration, and consolidates spend into one place so you can see cost per request across every model.
The trade-off is a thin layer of indirection and, usually, a small margin over raw provider cost in exchange for redundancy, caching, and unified billing. For teams using more than one or two models, the time saved typically outweighs the margin.
How Kyma does it
Kyma is an LLM gateway: one OpenAI- and Anthropic-compatible endpoint exposes open-source and frontier models for chat, reasoning, image, video, and audio, with multi-route redundancy and one balance. You switch models by changing the model field — see the model catalog and pricing.