Questions about your estimate

The estimate applies catalog rates to the activity and token assumptions you provide. Actual output length, conversation history, extra calls, discounts, taxes and services outside the modeled workload can change the bill. Inspect the quantities and formulas in the breakdown.

No. Consumer subscriptions and API billing are different products. ByeTokens estimates supported API usage, not the value or usage limits of a ChatGPT or Claude subscription.

Input tokens are the full billable input per call, including any conversation history, system messages or retrieved text you send. Calls per message or task multiply the workload. History growth and additional calls are not inferred automatically.

The current language-model adapter prices ordinary input and output tokens. It does not separately model cache read/write rates, batch discounts or reasoning-token billing. Automatic retries are excluded. If reasoning or another feature produces billable tokens, the provider invoice may differ from the ordinary text estimate.

Taxes, infrastructure, development costs, negotiated discounts and one-time promotional credits are excluded. Supported recurring free allowances are applied only when enabled. Supported fixed API subscription fees and included units are handled by their tariff rules; a monthly API plan is different from a consumer chat subscription.

It is the lowest exact total among complete compatible stacks with fresh prices in this catalog, under your filters and workload. It does not mean cheapest everywhere or best quality. Prices requiring another check can appear as stale alternatives but cannot receive the fresh Cheapest recommendation.

Verified describes the source and recency of a tariff check. It does not guarantee a future bill. Good fit describes task-specific quality evidence, using independent signals and uncertainty rules. A model can have fresh pricing without enough evidence for a fit recommendation.

A service may be catalog only, have an unsupported billing rule, lack current prices, exceed token limits or not meet your selected capabilities. Missing services are never treated as free. Data status describes catalog coverage; the calculator checks compatibility with your particular workload.

Saved scenarios include pricing and evidence versions. Current prices and evidence can change. The calculator identifies archived versions and offers current repricing where available. An old snapshot preserves its rates but does not make them current; rates older than 30 days cannot produce a current eligible estimate.