I know that GPT-5.5 xhigh isn’t available for Cursor Cloud Agents. Nevertheless from a user perspective without an explanation it is a bug.
Why are model selections limited in Cursor Cloud Agents?
Steps to Reproduce
They’re not there.
Expected Behavior
They should be there.
Operating System
MacOS
Version Information
Cloud Agent
For AI issues: which model did you use?
Well I wanted to use gpt-5.5 xhigh.
Additional Information
If I want to use gpt-5.5 xhigh and it’s unavailable in cloud agents then Cursor is unusable for the purposes of using that model with that reasoning effort.
Hey, thanks for the report. This isn’t a bug, it’s a current limitation. Cloud Agents use a curated server-side model list, and the web UI can’t pass reasoning effort as a separate parameter yet, like it does in the IDE with the new model picker. That’s why xhigh variants aren’t available in Cloud Agents right now.
We’re tracking this gap on our side, but I can’t share an exact ETA. The more upvotes these requests get, the more likely it is to be prioritized, so your report helps.
For now, the workaround is to run GPT-5.5 xhigh from the IDE, where the reasoning level selector is already available. If there’s an update on the cloud picker, I’ll post it here.
Technically yes, we can add more options to the current dropdown. That’s how we used to expand the list, for example GPT-5.5 high was added to Cloud Agents that way. But the direction we’re taking is to bring the same model picker from the IDE into the web UI, with a proper reasoning effort selector, so we don’t have to add separate items for every variant like xhigh, high, medium, and so on.
So we’re not planning a quick dropdown expansion right now. The focus is on the proper picker. I can’t share an ETA yet, but the request is on our radar. The more reports like this we get, the more likely it is we’ll pick it up sooner. I’ll post an update in the thread when I have one.
Hey Dean, is there an update on the ETA? I love gpt 5.5 xhigh. Cloud agents only with high is a noticeable drop in quality for my tasks. My workflow is kind of taking the hardest tasks & use local on xhigh, and cloud agents on high for the easier stuff, but that doesn’t scale well.
Idk what changed but just now I saw my cloud agents on my phone/browser have the model as “gpt-5.5-extra-high-fast”, despite not being in the dropdown. I started these from the Cursor desktop macos app with 5.5 xhigh fast mode. Looks like something is in the works? before, it was reverting to GPT 5.5 High when I started cloud agents from the desktop app
Hey, I can’t share an ETA for the native selector in the web dropdown yet. The request is being tracked, but there’s no timeline.
What you saw (gpt-5.5-extra-high-fast on a cloud agent started from desktop) is the working way to get xhigh on cloud right now. Start the agent from the IDE with GPT-5.5 xhigh fast selected, and the reasoning effort carries over to the cloud run even though the option isn’t in the web dropdown. So for your workflow, run heavy tasks from the IDE with xhigh instead of using the web dropdown.
I won’t claim this is an intentionally shipped feature. No guarantees, and the behavior might change. If you notice it switching back to High again, let me know. Once there’s an update on the native selector in the web UI, I’ll reply in the thread.
The reasoning level selector is now available on cursor.com/agents. The web model picker now works like the one in the IDE, so you can pick GPT-5.5 with xhigh reasoning when starting a cloud agent.
before, it was reverting to GPT 5.5 High when I started cloud agents from the desktop app