Use Zed / Continue with llmapi.cheap

Run Zed / Continue on Gemini 3.8 Flash, Claude Sonnet 4.6, Claude Opus 4.6 or GPT-OSS with a single llmapi.cheap key. Plans start at $2 a month, and there is a free one-hour trial.

Protocol: OpenAI Chat

Setup

Zed settings.json (set the LLMAPI_API_KEY environment variable to your key):

"language_models": {
  "openai_compatible": {
    "llmapi": {
      "api_url": "https://llmapi.cheap/v1",
      "available_models": [{ "name": "gemini-3.8-flash-high", "display_name": "Gemini 3.8 Flash (high)", "max_tokens": 1048576 }, { "name": "claude-sonnet-4-6", "display_name": "Claude Sonnet 4.6", "max_tokens": 200000 }]
    }
  }
}

Continue ~/.continue/config.yaml, under models:

  - name: llmapi Gemini 3.8 Flash (high)
    provider: openai
    model: gemini-3.8-flash-high
    apiBase: https://llmapi.cheap/v1
    apiKey: YOUR_API_KEY

Endpoints

Use the base URL that matches the protocol your tool speaks.

OpenAI: https://llmapi.cheap/v1
Anthropic: https://llmapi.cheap
Gemini: https://llmapi.cheap/v1beta
Models: https://llmapi.cheap/v1/models

Model IDs

Any of these work wherever Zed / Continue asks for a model name. Gemini 3.8 Flash (high) is the default the setup command picks.

Model ID	Model
`gemini-3.8-flash-high`	Gemini 3.8 Flash, high reasoning (default)
`gemini-3.8-flash`	Gemini 3.8 Flash, medium reasoning
`gemini-3.8-flash-low`	Gemini 3.8 Flash, low reasoning
`gemini-3.1-pro-high`	Gemini 3.1 Pro, high reasoning
`gemini-3.1-pro-low`	Gemini 3.1 Pro, low reasoning
`gemini-3-flash`	Gemini 3 Flash
`claude-sonnet-4-6`	Claude Sonnet 4.6
`claude-opus-4-6`	Claude Opus 4.6
`gpt-oss-120b`	GPT-OSS 120B

Troubleshooting

401 / invalid key: check the key starts with sk-llm- and has no extra spaces. Create a new one under Dashboard → API keys if needed.
Model not allowed: the free trial covers the Gemini models; Claude and GPT-OSS come with any paid plan.
404 on requests: OpenAI-style tools need https://llmapi.cheap/v1; Anthropic-style tools need https://llmapi.cheap without /v1.