Glossary: LLMs in Production

Effort

Also written: effort

A setting for how much a model thinks before answering. Lower effort means fewer output tokens and a faster first token, so Haiku tags chats at low effort.