# When using sonnet 3.7 massive Credit Drain **URL:** \ **Category:** Bug Reports\ **Created:** [February 25, 2025, 2:34pm UTC](https://forum.cursor.com/t/when-using-sonnet-3-7-massive-credit-drain/54796 "2025-02-25T14:34:24Z")\ **Posts on this page:** 6\ **Page:** 1
**Author:** ![Eddy](https://avatars.discourse-cdn.com/v4/letter/e/57b2e6/32.png) [@Eddy](https://forum.cursor.com/u/Eddy)\ **Post date:** [February 25, 2025, 2:34pm UTC](https://forum.cursor.com/t/when-using-sonnet-3-7-massive-credit-drain/54796/1 "2025-02-25T14:34:24Z")
When using the new Sonnet 3.7 in Agent mode ive noticed like many others it gets stuck with “Error connecting to anthropic” we click resume and it continues BUT Now every tool call (like running a test, when it fails fixing it) Costs a Credit so one Question to Sonnet 3.7 in Agent mode can run up to like 80-90 Cost in Credits like 6-7 Prompts and 500 Credits are gone When was this communicated?? wasnt it before 1 Credit for one prompt? I know one of our competitors whos Cascading (Now found it we are not allowed to name them in the forum as it blocks the message lol) charges for every Agent Action, but they give you 1500 actions ---
**Author:** ![intellect](https://sea3.discourse-cdn.com/cursor1/user_avatar/forum.cursor.com/intellect/32/33770_2.png) [@intellect](https://forum.cursor.com/u/intellect)\ **Post date:** [February 25, 2025, 4:38pm UTC](https://forum.cursor.com/t/when-using-sonnet-3-7-massive-credit-drain/54796/2 "2025-02-25T16:38:09Z")
Yes… and cascading works better than the cursor in many tasks… I hope this is a mistake. Because many people will switch to other tools if the cursor team changes the calculation algorithm. That would be very sad. ---
**Author:** ![wallynm](https://sea3.discourse-cdn.com/cursor1/user_avatar/forum.cursor.com/wallynm/32/20940_2.png) [@wallynm](https://forum.cursor.com/u/wallynm)\ **Post date:** [February 25, 2025, 7:56pm UTC](https://forum.cursor.com/t/when-using-sonnet-3-7-massive-credit-drain/54796/3 "2025-02-25T19:56:45Z")
Sadly i’ve experiencing same issue, the new version is draining my tokens. So sad, because the new model seems to be so nice. Even old sonnet models is draining my credits. > [@New version V0.47 burning too many tokens](https://forum.cursor.com/t/new-version-v0-47-burning-too-many-tokens/54428/2): > > I’m not sure why it’s burning so many tokens, but i’ve just runned a new agentic on claude-haiku, and the execution that should be cost 1/3 of a token, actually was 4 tokens, seems that the new Agentic mode is opening / seeing way more context, wich is very good. But its costing soo many tokens… Can’t handle this confused ---
**Author:** ![danperks](https://sea3.discourse-cdn.com/cursor1/user_avatar/forum.cursor.com/danperks/32/19087_2.png) [@danperks](https://forum.cursor.com/u/danperks)\ **Post date:** [February 26, 2025, 9:50am UTC](https://forum.cursor.com/t/when-using-sonnet-3-7-massive-credit-drain/54796/4 "2025-02-26T09:50:47Z")
> [@User aborted request. sonnet 3.7](https://forum.cursor.com/t/user-aborted-request-sonnet-3-7/54670/12): > > Hey, we are seeing an increase in failed requests due to our capacity issues with Claude 3.7 right now. This may cause both initial requests to fail, but also “resume” or “try again” options to fail. You may see errors regarding user requests or tool call failures during this time. To be transparent, during this time, failed Claude 3.7 requests may still count towards your usage. We are working to fix this logic shortly, and will look at retrospectively fixing usage for failures that have alre… ---
**Author:** ![Eddy](https://avatars.discourse-cdn.com/v4/letter/e/57b2e6/32.png) [@Eddy](https://forum.cursor.com/u/Eddy)\ **Post date:** [March 2, 2025, 12:34am UTC](https://forum.cursor.com/t/when-using-sonnet-3-7-massive-credit-drain/54796/5 "2025-03-02T00:34:17Z")
Sadly its not really a solution, atleast not to the issue that each Toolcall within cursor Agent eats a credit I can find it anywhere communicated, is that intentional or a bug? cause now the same issue is with running any other model, even had it running Gemini 2.0 Pro, it uses a toolcall, to read file 1 credit, then checks context uses another credit, then made change in two files, 2 credits then a test in the end, 2 credits for them when it made error another credit to fix like6-7 credits for one chat? ---
**Author:** ![system](https://sea3.discourse-cdn.com/cursor1/user_avatar/forum.cursor.com/system/32/99069_2.png) [@system](https://forum.cursor.com/u/system)\ **Post date:** [April 1, 2025, 12:35am UTC](https://forum.cursor.com/t/when-using-sonnet-3-7-massive-credit-drain/54796/6 "2025-04-01T00:35:03Z")
Closed on April 1, 2025, 12:35am UTC This topic was automatically closed 30 days after the last reply. New replies are no longer allowed.