VercelEngineering PostOfficial Source

AI Gateway adds unified fast mode support

AI Gateway now exposes one beta fast-mode option across models, letting coding agents request lower latency while retaining standard-speed fallback when no fast tier exists.

Vercel · Jul 29, 2026

Open Source Open Markdown Open JSON

Source Summary

AI Gateway’s **unified fast mode is in beta**. Setting speed to fast requests a model’s fast serving tier and falls back to standard speed when unavailable; a fast-specific slug addresses that variant directly.

Practical Implication

Keep the base model ID plus the speed option when graceful fallback matters. Use a fast slug in explicit routing or fallback lists; Claude Code users can toggle supported Anthropic Opus models with **/fast** after setup.

Agent-Ready Context

AI Gateway’s **unified fast mode is in beta**. Setting speed to fast requests a model’s fast serving tier and falls back to standard speed when unavailable; a fast-specific slug addresses that variant directly.

Keep the base model ID plus the speed option when graceful fallback matters. Use a fast slug in explicit routing or fallback lists; Claude Code users can toggle supported Anthropic Opus models with **/fast** after setup.

Fast tiers usually have a **higher per-token cost**, and model coverage is still growing. Requesting fast mode where no tier exists has no effect, so inspect returned routing metadata and pricing before assuming latency changed.

Connected Context · Feed7 Judgment

This generalizes fast serving from model-specific variants into a common routing control, but it is a preference rather than a latency guarantee: unsupported requests silently remain standard speed and fast tiers usually cost more. Against the prior routing signals, it makes response speed another explicit policy dimension that must be verified through returned route and pricing metadata.

Claude Opus 5 now available on AI GatewayOpus 5 is a concrete supported fast-mode route, while the unified option lets callers request that tier without replacing the base model ID and retain standard-speed fallback.Kimi K3 and Kimi K3 Fast with ZDR and US-based providers now on AI GatewayKimi’s separate fast endpoint illustrates the explicit-slug pattern; unified fast mode adds a common preference mechanism with graceful fallback across supported models.Routing rules now available on AI GatewayRouting rules can centrally rewrite or deny model choices, while fast mode adds a speed-tier request within that gateway-controlled selection policy.TRACE-ROUTER: Task-Consistent and Adaptive Online Routing for Agentic AITRACE-Router frames latency as part of task-level backend selection; unified fast mode adds a serving-tier knob, but its fallback behavior means actual routing metadata is needed to know whether that knob took effect.

Context Map

infracoding#gateways#coding-agents#model-selection

Uncertainty

Fast tiers usually have a **higher per-token cost**, and model coverage is still growing. Requesting fast mode where no tier exists has no effect, so inspect returned routing metadata and pricing before assuming latency changed.