Billing

Users pay through a credit balance. Every API request deducts from the balance based on input and output token counts.

Platform margin

An 8% platform margin is added on top of the requested top-up amount.

Requested: $10.00
Margin (8%): $0.80
Charged:    $10.80

Examples for small first payments:

Credit added	Platform margin	Amount charged
$5.00	$0.40	$5.40
$10.00	$0.80	$10.80

Note

We never modify model prices. The margin is applied only once, at top-up. Each API call deducts at the model's actual token price.

Token pricing

Prices are quoted per 1M tokens in USD. Some models use input cached, reasoning tokens, image, or audio second units.

Example price units:

Unit	Example value	Note
per 1M tokens	$5.00	Text input/output
per image	$0.002	Image generation
per second	$0.0008	Audio/video
per minute	$0.048	Realtime audio
per 1M token-hours	$3.20	Cache

Check the current price per model at Dashboard > Models.

Reasoning tokens

On reasoning/thinking models (e.g. Z.AI GLM, the Google Gemini thinking series) the reasoning tokens the model generates are billed as output tokens — even though they do not appear in the response text. A short answer can therefore consume more tokens than expected because of the reasoning produced behind it.

For transparency, the response usage block breaks this out:

{
  "usage": {
    "prompt_tokens": 20,
    "completion_tokens": 596,
    "total_tokens": 616,
    "completion_tokens_details": { "reasoning_tokens": 575 }
  }
}

completion_tokens already includes reasoning tokens; reasoning_tokens only shows how many went to reasoning. prompt_tokens + completion_tokens = total_tokens always holds.

To keep costs down, keep thinking opt-in: most thinking-capable GLM models default to off (Z.AI Thinking Control) and only reason with the :think suffix or reasoning: true. GLM-5.3 and GLM-5.3-Flash are exceptions: reasoning is always on, cannot be disabled, and plain requests run at low.

Balance and history

Dashboard > Billing shows:

• Live balance
• Top-up history (amount, margin, date)
• Per-request deductions
• Monthly spend summary

Top-up methods

The target dashboard method is:

• Credit card or another supported payment method through a secure checkout page.

Top-ups go through the secure checkout flow in the Dashboard. Cryptocurrency payment is not accepted.

Insufficient balance

If the balance cannot cover a request, the API returns 402 insufficient_balance. No tokens are consumed.

{
  "error": {
    "message": "Insufficient balance. Top up required.",
    "type": "insufficient_balance"
  }
}