Documentation

CheapAI proxies the major frontier-model families: Anthropic Claude (Opus, Sonnet, Haiku), OpenAI ChatGPT (GPT-5 family) and Google Gemini (2.5 Pro / Flash). The full catalog with up-to-date prices is on the Models page.

Use the canonical model ID — the same string you see on the Models page (the column «Model»). Pass it in the request body as the `model` field. No prefixes, no rewrites — we don't re-label models under aliases.

{
  "model": "claude-sonnet-4-6",
  "messages": [
    { "role": "user", "content": "Hello" }
  ]
}

You're charged exactly per usage tokens returned by the upstream provider — input, output, and cache (read / write) where applicable. No subscriptions, no minimums, no rounding. Cache read is usually charged at ×0.1 of the input rate, cache write at ×1.25 (depends on family).