One field now asks any model for its fast tier, and falls back when there is not one

Vercel's AI Gateway added a unified fast-mode abstraction, in beta. Set speed to fast and the gateway serves the fast tier where a model has one and falls back to standard speed where it does not, trading a higher per-token cost for lower latency. Shipped the same day: sign-in with ChatGPT as a Vercel authentication option, and purchasable additional custom environments for Pro and Enterprise teams.

What it means Latency-versus-cost stops being a per-provider rewrite and becomes one parameter - the practical version of the model-swappability people keep being sold.