Usage and Billing
How Ablatic meters usage and bills per API key
Billing is per API key, not per organisation. Each key carries its own billing model and its own counters.
Two billing models
A key uses one of two models:
slot: a monthly request allowance. The key gets a fixed number of requests per month.api: a monthly EUR budget. Usage is metered in money against a cap.
Unit and reset
Usage on api keys is metered in micro-euros. One micro-euro is one millionth of a euro, so spend is tracked at fine resolution and rounded to readable figures in the console.
Counters reset at Europe/Vienna midnight, and the reset is DST-aware. There is a daily reset and a monthly reset, so you can read both today's usage and the month to date.
Prompt caching
Talos caches the prefix of your prompts automatically, and your bill reflects it, the same way OpenAI and Anthropic do it.
- A cached input token, one that hit a warm prefix, is billed at 10 percent of the normal input rate. A 90 percent discount.
- Writing a prefix into the cache the first time it is seen costs the same as a normal input token. No write premium.
- Output tokens are unaffected.
On the OpenAI-style endpoint (/v1/chat/completions) caching is automatic, so you pay the cached rate on repeated prefixes without changing your code. On the Anthropic-style endpoint (/v1/messages) you mark cache points with cache_control, exactly as with Anthropic. The usage object reports the cached and freshly written token counts so you can see the split.
Cache entries live for about five minutes after each use. A conversation or an agent loop that keeps hitting the same system prompt and history pays the cached rate on that shared prefix.
Usage history
Usage is kept with a rolling history window of about 90 days. You can review recent days and the current month. Older detail rolls off, while a durable monthly archive is sealed for the record.
Console views
The console shows your usage and billing. You can see tokens spent, cost, and how close you are to your cap, broken out per day and per month.
Default budget
New accounts default to the api billing model with a EUR 50 monthly budget.
We set your budget per account. Plan against the current figure shown in your console, since that is the value applied to your key.