Allow custom models with same/similar IDs as built-in models, routed via Override Base URL

Ideas Feature Requests
1

I want to use my own OpenAI-compatible endpoint (BYOK + Override OpenAI Base URL) with a model ID that matches or closely matches a built-in Cursor model.

Example:

  • Built-in: Claude Sonnet 5.5 (claude-sonnet-5-5) → Cursor channel
  • Custom: claude-sonnet-5-5 → my Base URL (e.g. jiekou.ai / highwayapi)

Current behavior:

  • Adding claude-sonnet-5-5 fails with:
    The model "claude-sonnet-5-5" is already available as "Claude Sonnet 5.5".
  • Built-in Claude does not use Override OpenAI Base URL.
  • Only unrelated custom IDs can be added, so naming cannot stay consistent with the provider.

Expected behavior:

  1. Allow adding a custom model whose ID is the same as, or similar to, a built-in model.
  2. Keep them as separate entries, for example:
    • Claude Sonnet 5.5 (Cursor)
    • Claude Sonnet 5.5 (Custom / BYOK)
  3. Custom entry must use OpenAI API Key + Override Base URL.
  4. Built-in entry continues to use Cursor’s channel and subscription quota.

Why it matters: Providers often use IDs like claude-sonnet-5-5. Forcing a different name breaks compatibility and makes model selection confusing. Same name + different channel is the expected setup for BYOK / custom endpoints.

3

3 posts were merged into an existing topic: Cursor-managed models are routed through Override OpenAI Base URL