I’m not sure how much credit is available for Cursor Pro each month. How many tokens can be used for Auto or Grok 4.5, 4.6 or Composer in their individual versions? During these months, frequent updates by the authorities might have made me feel that the quota was constantly changing. For example, from August 19th to September 19th, in the first half of this month, I was using “auto” and felt that every 100 million tokens would increase by 10%. Now, with 370 million tokens, it’s only 50%. Could we provide users with a clear quota for usage? These months have made my usage very uncomfortable. I have to constantly check the bill to know exactly what percentage I have used. Instead of being able to clearly know how many tokens I can use for the A/B/C model. I know that the latest version “auto” will be distributed to different models based on the routing, but the bill and usage do not show exactly which model it has been transferred to. This has made me very anxious. I don’t know how much quota I still have available to use.I’m not sure how much credit is available for Cursor Pro each month. How many tokens can be used for Auto or Grok 4.5, 4.6 or Composer in their individual versions? During these months, frequent updates by the authorities might have made me feel that the quota was constantly changing. For example, from August 19th to September 19th, in the first half of this month, I was using “auto” and felt that every 100 million tokens would increase by 10%. Now, with 370 million tokens, it’s only 50%. Could we provide users with a clear quota for usage? These months have made my usage very uncomfortable. I have to constantly check the bill to know exactly what percentage I have used. Instead of being able to clearly know how many tokens I can use for the A/B/C model. I know that the latest version “auto” will be distributed to different models based on the routing, but the bill and usage do not show exactly which model it has been transferred to. This has made me very anxious. I don’t know how much quota I still have available to use.
I’m in the same boat as you. I started with the €20 plan, and it lasted until around the 20th, I think. I upgraded to Pro; the first month covered the whole month, but from the second month on, that plan didn’t even get me halfway through the month—and that was always on auto-mode. This last time, I set it to Composer with Sun & Moon on auto, and even then, it’s the 14th and I’m already at 99% usage. I know this is a business for Cursor, but it’s unacceptable that the value keeps dropping.
I also agree. Provide an estimated amount based on the corresponding model for the package usage. For example, for the pro+ package, using auto would allow approximately 2.5 billion Tokens. Using grok 4.6 high could allow for xx billion Tokens. This dynamic estimation enables us to use resources more systematically.
Hey @jie_yu, sorry this cycle has felt like a moving target. There’s no fixed “N tokens per model” number because included usage is measured by the cost of each request, not a token count, so the same tokens go much further on Composer 2.5 than on Grok 4.5/4.6 (rates on the Models & Pricing page; live totals on the Spending tab).
The shift you saw is from a change that landed mid-cycle: since late August, Auto bills at the rate of whichever model it routes to instead of a flat rate. Your cycle started Aug 19, so early days were on the old flat rate and everything since is at the routed rate, which is higher when Auto lands on Grok. We also raised the included usage for Cursor models around then.
Seeing which model Auto picked isn’t shown on the Usage page for individual plans yet, I’ve passed that along (plus @z_l’s per-model estimate idea), no timeline to share.
To make the rest of the cycle more predictable: pick Composer 2.5 directly (cheaper per token than Auto→Grok), start a new chat per task (long chats re-send everything each turn), and skip the Fast variants unless you need speed (~2x).
@Cocina_facil_con_jua if yours keeps climbing on Composer, the usual culprit is one long chat, a fresh chat per task is the biggest lever.
@mohitjain Thank you for your reply. You seem to be quite concerned about the voices of the users. But I wonder if you have noticed that in some posts, it is mentioned that Composer 2.5 has a significant regression in functionality? I didn’t mention this. It was because I wanted to know about the usage-related issues. Since I started purchasing the annual subscription package, the rates have changed every month (people who have actually used it would understand this better and would know that what I’m saying is true). This has made me worry every month about whether the credit limit would be enough to last until the end of the month. In the first half of the month, one billing method was used, and in the second half, another billing method was used. I hope to achieve uniformity and transparency as soon as possible.