Zed

Supported setup

In Zed's language_models.openai_compatible configuration, set capabilities.chat_completions to false for the model so Zed requests ${api_url}/responses. Set api_url to https://api.openmodel.ai/v1 without a trailing /.

Configure Zed

Add this provider to Zed's settings.json:

{
  "language_models": {
    "openai_compatible": {
      "openmodel": {
        "api_url": "https://api.openmodel.ai/v1",
        "available_models": [
          {
            "name": "gpt-5.4",
            "display_name": "OpenModel GPT-5.4",
            "max_tokens": 1050000,
            "max_output_tokens": 128000,
            "capabilities": {
              "chat_completions": false
            }
          }
        ]
      }
    }
  }
}

openai_compatible is Zed's required settings key. In this configuration, chat_completions: false makes Zed use POST /v1/responses; the key name is not a claim that OpenModel provides a generic compatibility layer.

Enter the API key through Zed's provider settings. For provider ID openmodel, Zed also reads OPENMODEL_API_KEY.

Before changing the example model or token limits, use Available Models as the source of truth and select a model whose supported_protocols includes responses.

Verify

Select the configured model and send a short message. A request to /v1/chat/completions means the model capability was not applied; OpenModel will reject that route.

Next steps

• Available Models
• Responses API
• Authentication