GPT-5.1 routes are liveExplore →
EveryRouter

Advanced

Prompt Cache

Preserve cache economics for long-context agents and repeated prompts.

Cache-Aware Routing

EveryRouter keeps session affinity where possible so provider switching does not destroy prompt-cache economics.

Cache metadata can be carried through headers and route decisions, and cache cost is separated from prompt and completion cost.

X-EveryRouter-Cache: enabled
X-EveryRouter-Session-Id: agent-session-42
X-EveryRouter-Cache-TTL: 3600

Invalidation

A new provider, incompatible model, policy change, or explicit clear instruction can invalidate cache affinity.

When cache behavior changes, the route metadata explains why the request moved.