The cursor client does not display advanced models such as GPT 5.5

Support Help
1

The cursor client does not display advanced models such as GPT 5.5, but the web version displays them normally. After I used up my credit limit last month, I used a custom model. After resetting the quota this month, it was found that gpt5.5 or ops models were not displayed, only grok, kimi, and glm were displayed.

4

Regional restrictions and cursor version issues have been ruled out

6

You can check the different subscription tiers on your billing page. Many high-tier models (like 1M High/Max) typically require a Pro+ or Ultra plan to appear.

8

My plan is Pro, and in June I was able to use GPT 5.5 normally on the desktop

10

Hey, thanks for the report. This isn’t a desktop catalog bug, it’s a regional model restriction. That’s why everything shows up on web, but not in the desktop app.

Anthropic, OpenAI, and Google limit access to some models in certain regions at the provider level. Cursor filters those models out of the desktop catalog based on the IP region the request comes from. So you only see models without those restrictions like Composer, Grok, Kimi, and GLM. The web app and Cloud Agents run from Cursor’s servers, not your local IP, so they show the full list. That’s why desktop vs web can look different on the same account.

The fact that GPT 5.5 worked on desktop in June also fits this. If Cursor traffic, including the model list request, went through a VPN with a non-CN exit IP, the filter didn’t trigger and the full list showed up. Your plan Pro or Free doesn’t matter here. The request region is what matters.

What you can do to get the full list back in the desktop app is make sure the catalog request goes through a supported region.

  1. Enable TUN mode or a system-wide proxy in your VPN so all Cursor traffic uses a non-CN exit IP.
  2. Alternative: force HTTP/1.1 via a local proxy. Add all four lines to settings.json at once, replace the port with your proxy port, then restart Cursor:
{
  "cursor.general.disableHttp2": true,
  "http.proxy": "http://127.0.0.1:7897",
  "http.proxySupport": "override",
  "http.proxyStrictSSL": false
}

Quick heads up. On versions 3.8+ there’s a separate local process that fetches the model list, and it uses its own HTTP/2 stack and doesn’t always pick up proxy settings, so option 2 can be flaky. This is a known limitation, I’ve shared it with the team, and I don’t have an ETA yet. The most reliable option right now is TUN mode.

More details and community findings are here: Desktop model catalog filtered since June 29, but web works (same account)

If you’re using your own OpenAI key (BYOK) and the model still doesn’t show up, tell me and we’ll check that as a separate case.