Some updates on pricing and tiers moving forward.

We've been blown away by the amount of love that has been shown to Windsurf so far, and usage has far surpassed our expectations. Under the hood, we’ve been providing users with uncapped usage while we figured out a proper path to a sustainable service. However, that resulted in hitting rate limits set by our upstream model providers.

We are rolling out new tiers and a proper credit system so that we can fairly allocate the capacity given to us while we simultaneously work to increase this capacity. This should immediately reduce resource_exhausted issues for Pro tier users. While this is not fun or ideal, we have been and will continue to pass along any cost savings from optimizations back to you.

Our goal at Codeium has always been to let every developer dream bigger. To build things bigger, better, and faster than they could before. And seeing what you all have built so far has gotten us more fired up to make the AI even better. Thanks for being part of this journey, and surf's up!

See the new tiering system at and get more details of how the credits work at

great to be transparent about this costing you a lot, but 500 credits is like less than a week's worth of prompts if you actually use it. and on ultimate tier "unlimited user prompts" doesn't mean much when flow action prompts are limited to 3000. pricing model needs more work...

We fully recognize that it may feel like there is little control over the premium Flow Action credit usage because there may be wide variability on the number of premium Flow Action credits used for each premium User Prompt credit. We have put in certain guardrails so that the AI cannot get in long loops of tool usage, and have spent an enormous amount of research work to align the steps of Cascade with the shortest path of analysis and actions to complete a self-consistent, complete unit of work. At the end of the day, our costs do scale with the number of premium model calls. We are committed to continuing to find ways to optimize our costs, and plan on passing these cost savings back to you in the form of more credits at the same price points. We want people to get as much value as possible from Cascade while being financially sustainable for us long term. Also, Cascade Base model!

going back to cursor

This price doesn't make much sense, TBH. It doesn't even match @cursor_ai current offer; It is significantly less value now. Cursor offers unlimited slow requests (with pro models), for just $5 more. You missed a huge opportunity here. Improvising is expensive.

@cursor_ai We have a Cascade Base model that you can use in an unlimited capacity! For Cascade, latency is critical so slow requests were not an option.

This is disappointing. I've already cancelled the subscription, deleted my account and uninstalled the editor.