Skip to content

Topic

Context budgeting

Deciding how many bytes or tokens of standing instructions a model loads before work begins, and what is deferred to files fetched only when a task matches.

Current clusters