Usage and spending

The Usage page shows the AI API requests you have sent, once each one’s charge is recorded. It shows totals for those requests (cost, savings, tokens, cache use and processing time) and details for your most recent requests. To open it, click Usage under AI in the console sidebar. The page, titled Log & Usage, includes requests from every API key in your account, including chats you send from the Playground. To see one key’s requests, use the API key filter.

​

Filter by date range, model and API key

• Date range: Last 7 days or Last 30 days (the default). The range covers whole calendar days, up to the current time today. Each request counts on the day it started.
• Model: All models, or one model you used in that range, even if it is no longer offered.
• API key: All API keys, or one key you used in that range, shown by its name and the start of the key. A revoked or expired key is marked (inactive).

Changing the date range resets Model to All models and API key to All API keys. If you choose both a model and a key, the page shows only that key’s requests to that model. The filters apply to the totals, the Log and every tab, except the spend forecast. The line under the filters repeats the range, your time zone, the model and, if you chose one, the API key’s name. Days and times use the Time Zone saved on your account. New accounts confirm it at first sign-in. If you never set one, it is UTC. You can change it in Settings. See Account settings.

​

Totals at the top of the page

Cache hit. Some models can reuse the start of a prompt they processed recently, such as a long system prompt. Cache hit is the share of your input tokens that the model read from this cache. Cached tokens are part of your input tokens, not extra tokens. They are billed at the model’s cache read price, or at the input price if the model has none. Storing a prompt in the cache is a cache write, not a hit. Request details show cache writes as standard and 1h (one-hour) writes. They are billed at the model’s standard and one-hour cache write prices, or at the input price if the model has none. A request that read nothing from the cache, including one on a model without prompt caching, shows a 0% cache hit. Each model’s prices are on Models and pricing. Latency. This is how long the model took to process a request. It is not the full round-trip time your application sees, and it is not the time to the first token. Requests without a recorded time are left out of the average. Latency depends on the model and on how many tokens the request reads and writes.

​

Log

The Log tab opens first. Its Recent requests table lists your 8 most recent requests that match the filters, with each request’s model, time, tokens, cost and latency. Latencies above 2,000 ms are highlighted. When token credits or a promotion lowered a request’s price, the Cost column also shows the full price struck through and the percentage saved. Click Show details on a row to see:

• Request ID: the value of the X-Request-Id header on the API response.
• Tokens: input, output, cache hit and cache writes.
• Status: always Settled, because the page shows only requests whose charge is recorded.
• Charge breakdown: shown when token credits or a promotion lowered the price. It lists the Full price, what Token credits covered (the amount and the number of tokens), the promotion and its discount, and the amount Charged.

​

Summary

The Summary tab charts your requests per day, with the Total requests, Daily average and Peak day. Open View daily request values to see each day’s count. Below the chart are short versions of the Change in spending and Spend forecast & balance runway cards described in the next section. Click View spending details to open the Spending tab.

​

Spending

​

Spend forecast and balance runway

This card covers your Entire account, so the filters don’t affect it. It shows your Average daily spend, Estimated month-end spend, Available balance (the same figure as Current Credits on the Billing page) and Estimated balance runway. The runway is how many days your balance would last at your average daily spend, with no top-ups or price changes. Below them, Charged this month shows your AI charges so far this month.

​

Why spending changed

This card compares the current period, up to now, with the previous period of the same length. It shows the Change in spending and splits it into estimated contributions that add up to the total. Covered by token credits shows what credits paid for. The charges on the card already exclude it. Largest changes by model lists up to five models.

​

Promotion savings

When promotions discounted your requests in the selected range, this card shows Total saved, Before discounts, Amount charged and Discounted requests, with the percentage saved. It has a row for each promotion, including promotions that have ended. Otherwise it shows No promotion savings in this period. This card’s Total saved and percentage count promotion discounts only, not token credits. Before discounts and Amount charged cover every request in the range, like the full price and Total charged at the top of the page. If token credits paid for part of your usage, the gap between them is larger than Total saved.

​

Models

The Model cost & performance table shows each model’s requests, tokens and cache hit, Total charged and Avg. duration (the average processing time). Click a model to see its Request share, Effective rate / 1M tokens, Cost / request and Tokens / request, its input and output tokens, and how many requests were measured for the duration. The effective rate is what you paid per 1M tokens after cache pricing, token credits and promotions. Because it combines input and output tokens, you can’t compare it directly with the separate input and output prices on Models and pricing. Use it to compare models with each other. To see what you saved, check the Total saved card, the request Charge breakdown, Covered by token credits and the Promotion savings card.

​

Tokens

The Tokens tab shows your input and output tokens and their share of all tokens, and your Average input / request and Average output / request next to the previous period. It also compares your cache hit with the previous period and counts your cache writes. The daily chart stacks Other input, Cached input and Output tokens. Click a legend item to show or hide it. Open View daily token values for each day’s exact counts. To see how input, cached, output and total tokens are counted, with worked examples, see How tokens are counted.

​

Reconcile usage with your balance

AI requests are charged after they finish. On the Billing page, the Transactions tab adds up each day’s AI charges into one AI usage transaction. It shows the date, the number of requests and the total amount. Click AI usage to open the Usage page. Days follow the Time Zone saved on your account. See Balance and billing. To check a period, compare Total charged (with All models and All API keys selected) with the sum of the AI usage transactions over the same days. The totals can differ slightly at the edges of the range:

• The Usage page dates a request by when it started. The AI usage transaction dates it by when its charge was recorded, so a request that runs past midnight can land on different days.
• A day whose AI requests cost nothing, for example because token credits paid for all of them, has no AI usage transaction.

Also keep in mind:

• VPS orders, renewals, top-ups and balance coupons change your balance but never appear on the Usage page. A coupon that unlocks an AI usage discount shows up as a promotion.
• A request that has already run is always charged, even if your balance goes below zero. Your next top-up pays off the negative amount first.
• To start a request on a paid model, your balance must be above zero, you must have token credits left for that model, or a promotion must cover the model’s full price.