Billing & Pricing

Nexotao uses a token-based system with a single Rupiah balance. There are no separate credits — all usage is drawn from the same balance.

Cost units

Everything is rounded up so small usage never becomes zero.

Cost headers

Every response includes X-Request-Id — the request id for tracking.

The cost header differs between non-streaming and streaming:

On a streaming response the headers are sent before the first token flows, so the final cost cannot be known yet. X-Cost-Rp is therefore deliberately omitted rather than sent as zero.

Getting the cost after a stream

Keep the X-Request-Id from the headers, then read the final figure from GET /usage/{request_id} once the stream has finished.

This endpoint uses a dashboard session, not an API key. It authenticates with your dashboard login token, so an sk-nexo-... key returns 401 auth.unauthorized. If you are calling from code that only holds an API key, read the breakdown on the Usage page in the dashboard instead.

The response carries the final, reconciled numbers:

{
  "model": "claude-opus-4-8",
  "input_tokens": 1240,
  "output_tokens": 380,
  "cache_read_tokens": 0,
  "cache_write_tokens": 0,
  "charge_micro": 16200000,
  "status": "ok",
  "turns": 1
}

• charge_micro — the cost in micro-rupiah (Rp 1 = 1,000,000 micro).
• turns — the number of charged upstream turns. Normally 1; it can be higher when the gateway ran a server-side tool loop, which is what explains a charge larger than a single turn’s tokens would suggest.

The usage record is written after the stream completes and is reconciled. Calling too early returns 404 with the code usage.not_found — wait a moment and retry. A 503 with dependency.database_unavailable is a transient failure; retry with backoff.

Live pricing

The table below is pulled directly from the live GET /models catalog (sell prices only):

Loading live pricing…

See also Models & Pricing for the full list, or nexotao.com/harga.