AI Gateway adds unified fast mode support
Vercel's AI Gateway now abstracts away provider choice—set speed: 'fast' and get lower latency across every model, with automatic fallback.
Why it matters
Practitioners building with multiple models can now treat latency/cost tradeoffs as a unified gateway setting rather than a per-provider configuration. Reduces ops friction for agents and real-time apps.
The key facts
17 to knowUnified fast mode abstraction across all AI Gateway models
Automatic fallback from fast tier to standard when unavailable
Fast mode available for growing set of models (including Claude Opus 5)
Fast variants trade higher per-token cost for lower latency/higher throughput
Feature works across all AI Gateway API formats
Claude Code integration with ./fast toggle for Opus models
Models without fast tier ignore the setting, run at standard speed
Published July 29, 2026 (beta feature)
Vercel AI Gateway unified fast mode now in beta
Single speed parameter works across all models on the platform
Automatic fallback from fast to standard tier if unavailable
Fast mode trades higher per-token cost for lower latency/throughput
Available for growing set of models; check pricing per variant
Works across all AI Gateway API formats
Claude Code toggle support for Anthropic Opus models
Fast variants typically more expensive than base models
Direct fast-slug addressing (e.g., anthropic/claude-opus-5-fast) also supported
Go to the source
Vercel Blogvercel.com
Publisher excerpt: AI Gateway has a unified fast mode abstraction, now in beta. You can now request the same way for every model on AI Gateway. Set speed to , and the gateway serves the fast tier when it's available and falls back to standard speed when it isn't.fast modefast Fast mode trades a higher per-token cost…