Windsurf shipped three updates at once: an Adaptive model router, a redesigned model picker with token pricing visible upfront, and the removal of daily quota limits for Max plan users. All three changes are live now for Pro, Max, and Teams subscribers.
Adaptive works as a routing layer inside the model picker. Instead of locking you into one model for every request, it dynamically selects the underlying model based on task complexity — routing a quick syntax lookup to a cheaper model and a multi-file refactor to something heavier. The quota draw is calculated at a fixed per-token rate regardless of which model gets selected. For the first two weeks, extra usage beyond quota is priced at $0.50 per 1M input tokens, $2.00 per 1M output tokens, and $0.10 per 1M cache read tokens.
The daily limit removal for Max users is a direct response to complaints about bursty workflows. Heavy users who front-load work at the start of a sprint were hitting daily caps before finishing a session. The weekly limit stays in place, but Max subscribers can now burn through it however they want. Windsurf kept daily limits on lower-tier plans intentionally — their data showed users on those plans risk exhausting a full week’s quota before learning how prompt caching affects costs.
TIP
Enable prompt caching awareness by watching the new cache timer in the context window indicator. A warm cache dramatically cuts token costs on long sessions — Windsurf now surfaces this directly in the UI so you can time your requests.
The updated model picker also shows token pricing per model inline, so you can see exactly what Claude Opus or GPT-5 will cost before sending a request. Response cards now include token counts after each message.
If you’re on Max and doing long coding sessions, the daily limit removal is the most immediately useful change. For everyone else, switching to Adaptive in the model picker is the lowest-effort way to extend monthly quota without changing how you work.