Read your usage and request logs

Usage tells you what your account was charged for. Logs shows each request behind the charges. Both cover everything made with your account: the chat, the voice input, and every call made with your API keys.

Usage#

Open Usage. For the period you choose, it shows:

• Totals for the period with a comparison against the one before: calls, spend and the average cost of a call.
• A chart by day.
• Breakdowns by function (model calls, web search and voice), by model, and by source, where the source is the chat or the name of the API key that made the call.

Filter by model, function, source or outcome, and widen or narrow the dates. Each total links to the matching requests in the log.

Logs#

Open Logs. Every request is one row, newest first, with:

• when it happened, the model and the function;
• the outcome: succeeded, failed or cancelled, and for a failure the reason in plain words;
• the quantity the model reported: tokens for chat and embeddings, images for images, seconds for speech;
• how long it took and what it cost, or Not charged, because a call that fails is not charged;
• its source.

The same filters apply. Open a row to see it in full.

Export#

Export CSV on the Logs page downloads one line per request for the current filters, for your own accounting.

Match it to your code#

A request made with an API key shows that key's name as its source, so a key per application tells you which application spent what. Each request has an id, which the API returns in the x-request-id header and the CSV export carries. That is how you find a particular call, or quote it to support.

What is not counted#

Nothing is estimated: a row shows what the model reported using, and the charge is for exactly that. See How calls are charged.