Where does the bug appear (feature/product)?
Cursor IDE
Describe the Bug
Hello. Adding custom models is not functioning correctly. The issue is that you are attempting to parse models from your database rather than obtaining them directly from the OpenAI Compatible API. Alternatively, your regular expressions may not be functioning properly.
For example, I add [Gemini CLI] gemini-2.5-pro or [Kiro] claude-opus-4-5-20251101, and they simply do not appear in the Models list in Settings. However, the model appears in the drop-down menu in the chat itself. Apparently, the metadata is not pulled from the database and a silent error occurs.
Because of this, the model cannot be removed from the settings. The model settings are not pulled, which breaks the token counters in the chat and prevents chat summarization from working. This cannot be fixed through the UI except by editing .vscdb.
This error exists in both stable and Nightly versions.
I believe that when Override Base URL is enabled, the entire list of available models from the API should be automatically added and users should be immediately given standard on/off toggles to control the models.
Or parse from the API and suggest which models to add to the list.
I would also like to add support for the Ollama API, it’s not difficult, this API is very similar to OpenAI. And possibly remove the restriction that prohibits the use of local Base URL addresses.
The screenshot shows two bugs, on the right the model was added but did not make it into the settings and remained only in the chat. It works but has problems with the description metadata that it could not obtain. It is impossible to delete it from the UI.
On the left, a model has been added to the list of models, but it has the opposite bug. It is enabled but cannot be activated in the chat or deleted from the settings.
For such cases, or until this behavior is stable, a button is needed to clear the list of custom models from the database.
Steps to Reproduce
The problem can be easily reproduced; just try adding a list of models yourself. Here is a list for testing.
[Antigravity] gemini-2.5-computer-use-preview-10-2025
[Antigravity] gemini-2.5-flash
[Antigravity] gemini-2.5-flash-image
[Antigravity] gemini-2.5-flash-lite
[Antigravity] gemini-3-pro-image-preview
[Antigravity] gemini-3-pro-preview
[Antigravity] gemini-claude-opus-4-5-thinking
[Antigravity] gemini-claude-sonnet-4-5
[Antigravity] gemini-claude-sonnet-4-5-thinking
[Antigravity] gpt-oss-120b-medium
[Cline] minimax/minimax-m2
[Cline] x-ai/grok-code-fast-1
[Gemini CLI] gemini-2.5-flash
[Gemini CLI] gemini-2.5-flash-lite
[Gemini CLI] gemini-2.5-pro
[Gemini CLI] gemini-3-pro-preview
[Kiro] amazonq-claude-3-7-sonnet-20250219
[Kiro] amazonq-claude-sonnet-4-20250514
[Kiro] claude-3-5-haiku-20241022
[Kiro] claude-3-5-sonnet-20241022
[Kiro] claude-3-7-sonnet-20250219
[Kiro] claude-4-sonnet
[Kiro] claude-haiku-4-5
[Kiro] claude-opus-4-20250514
[Kiro] claude-opus-4-5-20251101
[Kiro] claude-sonnet-4-20250514
[Kiro] claude-sonnet-4-5
[Kiro] claude-sonnet-4-5-20250929
[Qwen] qwen3-coder-flash
[Qwen] qwen3-coder-plus
[Qwen] vision-model
[iFlow] deepseek-r1
[iFlow] deepseek-v3
[iFlow] deepseek-v3.1
[iFlow] deepseek-v3.2
[iFlow] deepseek-v3.2-chat
[iFlow] glm-4.6
[iFlow] kimi-k2
[iFlow] kimi-k2-0905
[iFlow] kimi-k2-thinking
[iFlow] minimax-m2
[iFlow] qwen3-235b
[iFlow] qwen3-235b-a22b-instruct
[iFlow] qwen3-235b-a22b-thinking-2507
[iFlow] qwen3-32b
[iFlow] qwen3-coder-plus
[iFlow] qwen3-max
[iFlow] qwen3-max-preview
[iFlow] qwen3-vl-plus
[iFlow] tstars2.0
Operating System
Windows 10/11
Current Cursor Version (Menu → About Cursor → Copy)
Version: 2.2.0-pre.30.patch.0 (system setup)
VSCode Version: 1.105.1
Commit: f00523626e3ebe8a9bb0c606f576a52786677570
Date: 2025-12-07T08:20:56.874Z
Electron: 37.7.0
Chromium: 138.0.7204.251
Node.js: 22.20.0
V8: 13.8.258.32-electron.0
OS: Windows_NT x64 10.0.26100
Does this stop you from using Cursor
Yes - Cursor is unusable
