The token limit for Auto appears to have been reduced this month

I checked my usage history and found that my auto token limit was previously between 450-500M, but in my latest statement this month, I found that my auto token limit is only 300M. Has the limit been adjusted? Or is there some other reason?

The API limit remained unchanged, only the auto limit decreased significantly.

Hey. Auto doesn’t have a fixed token limit. Under the hood it works as a shared budget Auto plus Composer pool, separate from the API pool. The token count you see in usage history depends on which models the router picked this month, how well prompt caching worked, and those models’ per token prices.

So swings like 450 to 500M down to 300M month to month are normal and don’t mean the limit was cut. If routing went to more expensive models more often this month, or caching hit less, the same budget will produce fewer tokens.

More context on how the pool works: Models & Pricing | Cursor Docs and this thread Clarification on Cursor Pro vs Team Plan – Agent Usage and Token Allocation

Same here, i have record in oct-nov2025 that i already used 700M+ token, i always using auto mode, and i’ve never hit any limit no matter how long my prompt is, and i’am using it for several month everyday without issue (i’am always using $20 plan), but when i’am using it in around des 2025, i’am surprised because i hit the limit even in early week after plan activation, and now too, when i’am using only for about 5 days, the token used is already up to 140M, and it’s already labeled as used about 34% of my token, like, wtf? :face_with_raised_eyebrow:

Hey, I get that the numbers are confusing, but there isn’t a reduced limit here. Auto isn’t a fixed token count, it’s a shared dollar-budget pool (Auto + Composer). How many tokens it “fits” depends on which models the router picked this month, how well prompt caching worked, and those models’ per-token prices.

So the same budget can translate into a different token count month to month. If this month more requests went to more expensive models, or caching hit less often, you’ll “fit” fewer tokens and the usage percent will climb faster. That’s expected behavior, not a limit cut.

Also, comparing to Oct-Nov 2025 isn’t quite accurate. Pricing and how the pools work have changed since then, so you can’t directly match 700M back then to now.

More details on how the pool works:

If you think your specific account has a usage calculation bug, email [email protected] and they can check the account details.