I am assuming this is whats happening but i’ve been only using Grok 4.5 and Cursor 2.5. i did not use any of these other models and seeing SONNET 5 and gemini 3.1 pro?
I dont know whats going on, but thats ALOT of tokens spent for models i havent touched. Is there any way to fix this? I am using cursor-agent and i have the latest version. i dont even want to use cursor anymore because i dont want work done by sonnet5 or gemini 3.1 pro.
Hey, thanks for the report. From the screenshots I can see usage for Sonnet 5 and Gemini 3.1 Pro, even though you only selected Grok 4.5 and Composer 2.5.
What’s happening: when the main agent delegates part of the work to subagents via the Task tool, the subagent can pick a different model than the one you chose for the session, sometimes a more expensive one. This isn’t anything in your setup, and you didn’t do anything wrong. It’s a known issue we’re tracking. I can’t share an exact ETA yet, but I’ll update the thread when there’s news.
A few things that might help right now:
cursor-agent, there isn’t a full model catalog for subagents yet, so you can’t directly pin the subagent model there.subagentStart hook Hooks | Cursor Docs. It lets you allow or block subagent creation, including by model.Related threads if you want to follow along:
Separately about quota usage: if you want us to check your specific account, what was charged and what we can do about it, email [email protected]. Our accounts team handles that. Let me know if you still have questions about the subagent behavior.
This is happening to me as well. In my current settting, I eliminated all of the rest of model choices except for Grok 4.6. So supposedly, all of my models, as well as, subagents should use Grok 4.6. Below is a screenshot where it shows that when the VM is spawning subagents for computer usage, it uses Sonnet 4.5. This is a bit annoying in the sense that I have exhausted my “Other Models” usage, and I disabled on-demand usage previously. Yesterday I had to enable on-demand usage and pay extra so that Cursor unblock my automation and my cloud agent usage.
I also attached an image of my usage where Sonnet 4.5 shows up on date Aug 23rd.
I feel like I should not be paying for Sonnet 4.5 as it is extra cost outside my current plan.
Hope that it can be resolved soon enough!
To help solve this issue, I can provide some more details. It seems that there are different subagents category, there are the ones for “Computer Use”, and there are subagents categoried as “General Purpose,” and perhaps there are other categories. As far as my observation goes, it is always the “Computer Use” that are creating troubles!
I could be wrong, but just some insight.
Hi Rie, sorry to bother, but I wrote an email to the account team, but then there is this automated Sam AI assistant who is not aware of the situation. I would like to speak to a human representative to perhaps resolve this issue.
This is my ticket number: T-F7259, let me know if you can offer some assistance.
Hey @Junyi_Zhou. Your screenshots make it clear what happened. Cloud Agent launched the Computer Use subagent on Claude Sonnet 4.5 even though you only left Grok 4.6 enabled, and it showed up under Other Models.
Your point about Computer Use vs General Purpose is spot on. Specialized subagents like Computer Use are currently tied to specific models and don’t yet respect the model restrictions you set. So this isn’t something you misconfigured and it’s not a mistake on your end. It’s the same issue mentioned earlier in the thread and we’re tracking it. I can’t share an exact ETA yet, but I’ll post here when there’s an update.
On billing, decisions about the specific charge and any possible adjustment are handled by the Accounts team. You already have ticket T-F7259 in their queue. A human will review it, the automated reply isn’t the final answer. No need to repost billing details here, the right team already has everything they need.
Let me know if you still have questions specifically about subagent behavior.
