I can add a model with any random non-existent name without issue, but I fail to add the GPT-5.5 or claude-opus-4-8 model.The API Base URL is a third-party website address, not the official OpenAI URL.
Hey, thanks for the report. This isn’t a bug, it’s expected behavior. I’ll explain how to do what you need.
gpt-5.5 and claude-opus-4-8 are built-in model IDs in Cursor. When you try to add a custom model with the same name, Cursor detects a collision and won’t let you create a duplicate. That’s why you see No models available. Random names like 1234 don’t match any built-ins, so they get added fine.
To run gpt-5.5 through your third-party endpoint, don’t add it as a custom model. Do this instead:
- In Settings → Models, enable OpenAI API Key and Override OpenAI Base URL, then set your URL and key.
- Use the built-in GPT-5.5, and the request will go to your endpoint.
Important: Cursor doesn’t change case and doesn’t rename the model. Your endpoint will receive exactly gpt-5.5 in lowercase as the internal ID. GPT-5.5 is just the display name in the UI. So if your API is case-sensitive, it will still get gpt-5.5.
About claude-opus-4-8. That’s an Anthropic model, and Override OpenAI Base URL won’t route it since it only works for the OpenAI family. For BYOK Claude, use the Anthropic API Key section. That’s what routes all claude- models.
This is a known limitation, and I can’t share an ETA for a change yet. If you hit a payload format error on /chat/completions after setting the override, that’s a separate known issue. Send the Request ID and we’ll take a look.