Invalid URL when calling OpenRouter model (`prime-intellect/intellect-3`, `deepseek/deepseek-r1-0528`)

[DamianB-BitFlipper]

Description

Description

I’m getting a TypeError [ERR_INVALID_URL]: "undefined/chat/completions" cannot be parsed as a URL when trying to run inference against an OpenRouter-hosted model (prime-intellect/intellect-3, deepseek/deepseek-r1-0528). The failure occurs even though an OpenRouter API key is correctly configured and funded and /connect reports the provider as active. I am able to inference the default provided openrouter models.

The error suggests the client is not resolving the OpenRouter base URL, resulting in "undefined/chat/completions" being passed to the fetch call.

Config

Follows the opencode.ai docs

"provider": {
  "openrouter": {
    "models": {
      "prime-intellect/intellect-3": {},
      "deepseek/deepseek-r1-0528": {}
    }
  }
}

OpenCode version

1.0.134

Steps to reproduce

1. Add the providers in the config above to the opencode.json
2. Connect the openrouter provider with an API key
3. Try to inference, for example opencode run -m openrouter/prime-intellect/intellect-3 hello

Screenshot and/or share link

No response

Operating System

macOS Sequoia 15.5

Terminal

kitty

[github-actions[bot]]

This issue might be a duplicate of existing issues. Please check:

• #195: Attempting to use LM Studio with OpenCode - Same exact error message "TypeError [ERR_INVALID_URL]: "undefined/chat/completions" cannot be parsed as URL" when trying to use a custom provider with baseURL configuration

The root cause appears to be related to how custom provider baseURL configurations are being resolved. Issue #195 also mentions using OpenAI-compatible providers with baseURL, which is similar to your OpenRouter setup.

Feel free to ignore if your specific case differs.

[rekram1-node]

fixed in next release: https://github.com/sst/opencode/commit/c30b1130eea1d2317105b4b3e3f69cfeb5bdc6de

[DamianB-BitFlipper]

Amazing great. Thanks!