https://api.model.sale/v1 works with standard OpenAI-compatible clients.ChatGPT API · prepaid developer access
A simple ChatGPT-compatible API.
Use one Model.sale key with the OpenAI SDK, Codex, curl or your own backend. No monthly subscription: add balance when you need it and pay for measured usage.
How to make your first request
- Create an account and generate a Model.sale API key. The full secret is shown once.
- Add at least $5, or use the one-time Telegram test credit when eligible.
- Choose an ID marked LIVE in the catalog.
- Set the base URL and key in your server environment, then verify the request in Usage.
export OPENAI_BASE_URL=https://api.model.sale/v1
export OPENAI_API_KEY=ms_live_...
curl "$OPENAI_BASE_URL/responses" \
-H "Authorization: Bearer $OPENAI_API_KEY" \
-H "Content-Type: application/json" \
-d '{"model":"gpt-5.6-luna","input":"Reply exactly OK"}'Keep the key in a secret manager or local environment file. Never put it in a URL, browser bundle, shell argument, repository or analytics event.
Responses
Recommended for Codex and GPT models that expose /v1/responses. Streaming ends with a terminal response event.
Chat Completions
Use /v1/chat/completions for compatible GPT, GLM and DeepSeek rows. Set stream: true for SSE.
Exact settlement
The wallet reserves funds before dispatch and settles terminal input/output usage. The unused reservation is released.
What “ChatGPT-compatible” means here
Model.sale provides a developer API, not the consumer ChatGPT product. Existing OpenAI SDKs can use standard request shapes, but each model still has its own validated capabilities, limits and endpoint list. The requested model ID is preserved; an unavailable model is never silently replaced.
| Need | Use |
|---|---|
| GPT or Codex integration | Responses where the model row lists it |
| GLM or DeepSeek | Chat Completions and SSE where listed |
| Current model IDs and prices | Full catalog and pricing |
| Client setup | Guides for Codex, SDKs and Open IDEs |
Billing and privacy
Rates are blended charges per 1M charged tokens; cached and reasoning fields remain visible when the protocol reports them. Missing terminal usage stays pending for reconciliation instead of being guessed. API prompts, source code and response bodies are not stored by default.
Read the pricing methodology, API documentation and reliability snapshot before production traffic.
FAQ
Can I use my regular OpenAI key?
No. Create a Model.sale key and set the Model.sale base URL.
Do I need a subscription?
No. Model.sale is prepaid. The minimum deposit is $5 and there is no monthly seat fee.
Where can I see the exact charge?
Every completed request appears in Usage with model, endpoint, token fields, request ID and the settled charge.