Caching and refunds

Repeats cost half. Unused budget comes back.

Lane cache

A byte-identical request can be answered from cache without reaching a model. A hit is billed at 50% of the normal token charge and returns X-Inferlane-Cache: HIT.

• Deterministic by default. Only requests with temperature: 0 are cached automatically.
• Per-request control. Send X-Inferlane-Cache: on to cache at any temperature, or off to bypass.
• Scoped to you. The cache is keyed to your key or paying wallet. A hit never crosses accounts.
• Non-streaming only.
• Fails open. Any cache problem simply sends the request upstream.

curl https://inferlane.xyz/v1/chat/completions \
  -H "Authorization: Bearer $INFERLANE_KEY" \
  -H "X-Inferlane-Cache: on" \
  -H "content-type: application/json" \
  -d '{"model":"meta-llama/llama-3.3-70b-instruct","messages":[{"role":"user","content":"Define a rate card."}]}'

Refunds

Value is owed back to a wallet in four cases:

• Unused ceiling. A pay-per-call ceiling minus the real charge.
• Failed call. You paid and no upstream answered. All of it returns.
• Failed deposit. A deposit settled but could not be credited to a key.
• Closed key. The remaining balance of a revoked key.

Refunds are grouped per wallet and sent on Robinhood Chain in a regular sweep. Very small amounts wait until they add up to the minimum transfer.

Empty replies are free

If a model returns no usable text, the charge is waived. On pay-per-call the full ceiling is refunded.