API Documentation

01 · QUICKSTART

Make your first request

Create an API key in the dashboard, select the right group and point your client at https://orbiqen.com/v1. Do not append /v1 twice.

Add balance. The service uses prepaid USD credit.
Create a key. Choose its group and permitted models before sharing it with an app.
Test the connection. Send a short request and inspect it in Usage.

OpenAI-compatible APIPOST /v1/chat/completions

curl https://orbiqen.com/v1/chat/completions \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"gpt-5.5","messages":[{"role":"user","content":"Hello"}]}'

02 · ACCESS CONTROL

API keys and routing groups

A customer API key does not inherit the full administrative catalog. It can only use the models enabled in the group selected when that key was created.

Group	What it enables	What it does not enable
Economy ChatGPT	Selected GPT models for common tasks.	Claude and stable image models.
Stable ChatGPT	Selected GPT models and published image models.	Claude, unless you create a Claude key.
Claude	Only the Claude models enabled for that key.	GPT, image and Chinese models.
Chinese models	Published models from that family, model by model.	GPT and Claude, unless another key is used.

Best practice:use a separate key for each application, customer or environment. Set a spending cap and revoke keys that are no longer needed.

03 · TRANSPARENT BILLING

How a request is billed

Published rates are final USD prices per million tokens. A request can include input, output and, when reported by the model, cache reads or writes.

(input × input rate) + (output × output rate) + (cache × cache rate)

What you will see

Input: tokens sent to the model.
Output: tokens generated by the model.
Cache read/write: reused context when available.

What is not added

No required subscription.
No fixed per-request fee.
No hidden markup when balance is deducted.

Prices can differ from official providers or other aggregators because capacity, routes and commercial conditions differ. Orbiqen does not promise to always be the cheapest option: the active rate is published and every request remains auditable.

04 · AVAILABILITY

Internal routes and fallback

Primary route

This is the preferred connection for a group and model. Customers always use the same public Orbiqen endpoint.

Fallback

It can be used when an upstream route is unavailable. It does not disclose internal providers and is not a fixed-latency guarantee.

Stable billing:the price of your request depends on its model and API key group, not on the internal route used to fulfill it.

05 · AUDIT

Verify every charge in Usage

After making a request, open Usage in the dashboard. The history gives you enough detail to explain a deduction without relying on estimates.

Model and groupConfirm which model processed the request.

TokensSeparate input, output and cache when available.

CostCompare the balance deduction with the published rate.

Duration and statusFind slow responses, failures or retries.

06 · TROUBLESHOOTING

Common errors

Message	What to check
`401` or invalid token	The key is incomplete, revoked or sent using the wrong header.
`403` model access denied	The model is not enabled for that key or belongs to another group.
`400 Invalid URL`	Use `https://orbiqen.com/v1` as the Base URL and do not duplicate `/v1`.
Insufficient balance	Add credit or reduce the output limit for your request.

07 · ANTHROPIC COMPATIBILITY

Use Claude with the native API

Keys from the Claude group can also use Anthropic's native message format. Keep the model within the set allowed for that key.

Anthropic-compatible APIPOST /v1/messages

curl https://orbiqen.com/v1/messages \
  -H "x-api-key: YOUR_API_KEY" \
  -H "anthropic-version: 2023-06-01" \
  -H "content-type: application/json" \
  -d '{"model":"claude-sonnet-5","max_tokens":256,"messages":[{"role":"user","content":"Hello"}]}'

08 · FAQ

Before you go to production

Can I use one key for everything?

You can, but separate keys per project make it easier to revoke access, set limits and identify usage.

Do image models need another endpoint?

Published image models use the compatible Orbiqen route. Enable them only on the stable keys that need them.

Does fallback activate when a request is slow?

Fallback is for route errors or unavailability. It should not be treated as a timeout mechanism for requests that are still running.

How can I test with US$ 1?

Send short requests, choose the right model for the task and inspect Usage after every test. Actual cost depends on context and generated tokens.

Documentation to integrate, measure and control your usage.