Max mode allows the model to reach its maximum possible context window (Claude cursors are typically limited to 200k context windows). Enabling max mode allows the context window to reach 1m. Secondly, it allows you to call premium-related sub-proxies (of course, enabling max mode increases the cost by 1.2).
You didn’t understand my question. My question is, why can the thinking mode run in non-max mode, while the less costly non-thinking mode can only run in max mode?
It’s quite fun that my claude-opus-4-7-thinking-high couldn’t run in non-max mode now, but a few hours ago I could run it.
Just found out about this as well. Weird right?
Apparently, only opus 4.7 extra high thinking is allowed for non max in my IDE. Could be a bug though.