If you’ve been juggling 4 model picks a day to keep your Windsurf quota alive, that decision just got taken away from you — and made cheaper.
Windsurf shipped three changes at once aimed at making their token-based pricing less painful.
What “Adaptive” actually does
Adaptive sits in the model picker as a selectable option. Instead of you choosing a specific model, it dynamically routes each task to an appropriate underlying model — the idea being that simple tasks don’t burn premium-model tokens.
The updated model picker also integrates a prompt cache timer into the context window indicator and adds token counts to response cards, so you can see exactly how each message cost was calculated.
TIP
Prompt caching is a critical factor in request costs. Watch the new cache timer — letting the cache expire between messages can significantly increase your token spend.
Pricing for extra usage
For usage beyond your included quota, Windsurf is offering introductory rates for the first two weeks:
What Max users get
The daily limit is gone. The weekly limit stays, but Max users can burn through it on a single intense day if their workflow demands it. Windsurf kept daily caps on other plans deliberately — their data showed it’s too easy to exhaust a full week’s quota before learning to manage prompt caching and model selection.
NOTE
Windsurf says they’re building a “more efficient harness” with multi-model architecture and subagents to deliver higher quality at lower cost. No timeline given.