Cursor models are now MAX mode only

Context: Cursor made an overnight change for enterprise users on “Legacy Pricing” (request-based) that now forces all premium models—every Claude model and GPT models above 5.2—into Max mode. In practice, a single Claude 4.6 prompt that previously cost 2 requests can now cost 50. Many companies only give users around 500 requests total.

The economics may not have been working for them, but the way they handled this has seriously undermined my trust. Threads discussing the change across Reddit and the Cursor Forum have been deleted or locked.