Provide support to connect to an OpenRouter preset / specific model by chatLanguageModels.json

[Spitfire1900]

OpenRouter now supports presets, which can be used to fine tune model access. Such as requesting GLM-4.5 Air (free) and falling through to GLM 4.7 Flash when free inference is limited.

OpenCode can connect to presets via config like the following

{
  "provider": {
   "openrouter": {
      "npm": "@ai-sdk/openai-compatible",
      "name": "OpenRouter (external)",
      "options": {
        "baseURL": "https://openrouter.ai/api/v1"
      },
      "models": {
        "@preset/free-then-cheap": {
          "name": "Free then Cheap (OpenRouter)"
        }
      }
    }

  } 
}

However OpenRouter does not list presets in calls to https://openrouter.ai/api/v1/models, even with the appropriate Auth token, so they have to selected manually.

Provide support to list list in ~/.config/Code/User/chatLanguageModels.json, e.g.

[
	{
		"name": "OpenRouter",
		"vendor": "openrouter",
		"apiKey": "<api_key>",
		"models": [
			{
				"id": "@preset/free-then-cheap",
				"name": "Free then Cheap",
				"url": "https://openrouter.ai/api/v1/chat/completions",
				"toolCalling": true,
				"vision": false,
				"maxInputTokens": 128000,
				"maxOutputTokens": 16000
			}
		]
	},
]

[vs-code-engineering]

We closed this issue because we don't plan to address it in the foreseeable future. If you disagree and feel that this issue is crucial: we are happy to listen and to reconsider.

If you wonder what we are up to, please see our roadmap and issue reporting guidelines.

Thanks for your understanding, and happy coding!

[ZaituDev]

Manually creating @preset/slug strings on the web UI and copying them into config files creates unnecessary friction when setting up CLI tools. Having direct terminal controls makes switching routing configs much faster. Check out OpenPreset, a zero-dependency POSIX shell utility that lets you inspect provider health and manage OpenRouter presets right from your command line.