Skip to main content

PING_OUTPUT_ESTIMATE

Constant PING_OUTPUT_ESTIMATE 

Source
pub const PING_OUTPUT_ESTIMATE: u64 = 256;
Expand description

Expected output of a ping under budgeted thinking: a short thinking block plus “ok”. The API forces max_tokens to budget + 1 (≥ 1,025) there, and the budget is part of the message cache key, so the ping cannot ask for less; it is told it needs no reasoning and in practice thinks a sentence. 256 leaves several times that as headroom. Only the start decision uses it: the hard bound still prices every ping at max_tokens (see Period::bound_usd).