Usage and quota

Reports how much of Token Factory you’ve used, and what’s left of today’s allowance. This is the same data as Credits & Usage in the console.

GET /v1/usage API key

Request

Parameter	Type		Description
include_recent	boolean	Optional	Include recent individual requests. Defaults to false.

curl "https://developer.amd.com.cn/radeon/api/v1/usage?include_recent=true" \
  -H "Authorization: Bearer $RADEON_API_KEY"

Response

{
  "rpm_limit": 30,
  "daily_cost_limit_usd": 10,
  "daily_cost_used_usd": 1.8412,
  "daily_cost_remaining_usd": 8.1588,
  "daily_reset_timezone": "Asia/Shanghai",
  "daily_reset_at": "2026-08-26T00:00:00+08:00",
  "today": {
    "requests": 142,
    "errors": 3,
    "total_tokens": 88214,
    "prompt_tokens": 31002,
    "completion_tokens": 57212,
    "cost": 1.8412,
    "last_request_at": "2026-08-25T14:22:07+08:00"
  },
  "last_30_days": { "requests": 2840, "total_tokens": 1904221, "cost": 39.77 },
  "all_time":     { "requests": 9120, "total_tokens": 6210443, "cost": 128.44 },
  "by_model": [
    { "model": "DeepSeek-V4-Flash", "requests": 120, "total_tokens": 74001, "cost": 1.55 }
  ],
  "recent": []
}

Fields

Field	Description
rpm_limit	Requests per minute allowed for your key.
daily_cost_limit_usd	Today’s usage cap.
daily_cost_used_usd	Used so far today.
daily_cost_remaining_usd	What’s left before requests are refused.
daily_reset_timezone, daily_reset_at	When the daily allowance resets: 00:00 Beijing time.
today, last_30_days, all_time	Totals for each period.
by_model	The same totals per model.
recent	Recent individual requests. Empty unless include_recent=true.

Usage and cost values are in Token Factory points (pts), the unit shown on Credits & Usage. The field names end in _usd, but you aren’t charged money, and these points have nothing to do with your credits.

Token counts are split into prompt_tokens and completion_tokens. Some models also report reasoning_tokens.

Limits differ between accounts

The numbers above are an example. Read your own limits from this endpoint rather than assuming them, because they differ between accounts and change over time.

Monitor usage

Check before a large batch job that there’s enough left, or watch daily_cost_remaining_usd so a job doesn’t stop halfway. Once the allowance is used up, requests are refused until daily_reset_at.

Every Token Factory response also carries the current counters as headers. See Rate limits.

From the console

The Credits & Usage page reads the same data from GET /api/profile/model-usage, which takes a browser session instead of an API key. It adds a status field: ok, not_configured if the account has no API key yet, or not_available if there’s no usage data. A 502 from it means the usage service can’t be reached; treat that as unknown, not as zero usage.