AI Gateway adds unified fast mode support

**AI Gateway** from Vercel now offers a **unified fast mode** abstraction, currently in beta. You can request fast mode the same way for every model on AI Gateway instead of handling speed options per model.

Set **speed** to **fast**, and the gateway serves the fast tier when that tier is available for the chosen model. When the fast tier is not available, it falls back to standard speed. The abstraction is the same across models, so you do not pin to a specific provider’s fast endpoint or maintain separate call paths.

Fast mode trades a higher per-token cost for lower latency or higher throughput. The gateway applies that tradeoff only where the fast tier exists; elsewhere you still get a valid completion at standard speed without changing the request shape.

For engineers and builders, this matters because multi-model apps often mix providers and tiers. A single **speed=fast** setting keeps client code, routing, and retries simpler when you want the low-latency path without rewriting per-model branches.

In a market where gateways compete on routing, fallbacks, and cost–latency controls, a unified fast mode reduces the need to hardcode which models support a fast tier. You get fast mode wherever it is available, without pinning to a single model or vendor path.

Practical next step: try **speed=fast** on your existing AI Gateway calls, measure latency and token cost against standard, and watch how often the gateway actually hits the fast tier versus falling back—especially for models you rely on in production.