alt Hacker NewsDon't they consume less of the token quota in case the subagents are running cheaper models like Sonnet and Haiku compared to Opus?
Don't they consume less of the token quota in case the subagents are running cheaper models like Sonnet and Haiku compared to Opus?
Correct—I just wouldn't want folks to mistakenly think that the context fill % corresponds 1:1 with session token use.