Cost Estimates

Why the inference cost is shown as a range, and what you're charged.

Customer-facing cost and savings figures are ranges, stamped with the date of the catalog prices used.

Why a range

Before execution, output length and final provider usage are unknown. PeerLM therefore presents 0.5×–2× bounds around the estimate instead of false precision. After execution, receipts use measured usage where available and keep the same dated-price contract.

When no savings appear

Savings are absent, never zero, when observed cost cannot be resolved from a recorded cost or complete token counts and dated catalog prices. Latency is also omitted unless both sides share the same known measurement scope.

What PeerLM charges

Paid plans include inference within their hosted-generation ceiling. The customer unit is a Run, not tokens or an internal reservation amount. Enterprise can configure provider keys; the ceiling is exempted only when every provider PeerLM would fund for generations and judges has a successfully decrypted customer key.