Skip to main content

Consumption and usage

Every license key has a Consumption view in the Console that shows the token usage your deployment has reported against that license, a month-to-date summary with an estimated cost, and a pricing timeline that breaks the cost down rate window by rate window. This is where you go to understand what drove your token spend and how it maps to a bill, per deployment. Billing across all your keys is aggregated separately on the Billing page.

Usage is metered on prompt (input) tokens, the same measure billing uses (see Billing). None of these views show your prompts, documents, or results — only counts of tokens, attributed to the license.

The Consumption view​

Open a license key and go to Consumption to see the usage that deployment has reported over time. The chart plots prompt tokens consumed, and you can shape it three ways:

  • Choose a period — Last 24h, Week to date, Month to date (the default), or Last 30 days.
  • Choose the granularity — view usage in hourly or daily buckets. Hourly is useful for pinpointing a burst of activity; daily is easier for stepping back to the bigger picture over a longer period.
  • Break usage down — group consumption by model (the model that produced the tokens) or by activity (the kind of work — for example a chat turn, a workflow step, or the analysis coder's own model calls). The chart stacks the breakdown so you can see which models and which kinds of work drive your token spend.

Because the breakdown is attributable, you can answer questions like "which model is most of my spend on this month?" or "was yesterday's spike chat or a scheduled workflow?" directly from the chart.

Month-to-date summary​

Above the chart, a Month-to-date summary gives you the running total for the current billing month at a glance:

  • Prompt tokens consumed so far this month against this license.
  • Estimated cost for those tokens.

The month-to-date figure is an estimate that updates as new usage arrives — it is not an invoice. Your actual invoice is issued on the 1st of the following month for the previous month's usage, and once issued its figures are fixed (see Billing).

Pricing timeline and per-segment cost breakdown​

A single month can span more than one rate. The Pricing timeline shows the rate windows that applied during the period and breaks the estimated cost into one segment per window, so the total is always traceable back to the rate that produced it.

Each segment shows:

tokens used  @  rate  =  subtotal

and the period total is the sum of all the segments' subtotals:

total = subtotal(segment 1) + subtotal(segment 2) + …

For example, if a rate changed mid-month, you would see two segments — the tokens consumed before the change costed at the old rate, and the tokens after it costed at the new rate — and the total is their sum. This makes it easy to reconcile a month that straddled a price change: you can see exactly how many tokens fell into each window and what each window contributed.

Evaluation periods show $0. If a window is a free evaluation period, its segment counts the tokens you used but shows a $0 subtotal, so those tokens are visible in your consumption without contributing to the cost total.

How rates apply​

  • Rates are day-granular. A rate window starts and ends on whole-day boundaries, so a given day's usage is always priced at a single rate.
  • Rates are non-retrospective. A rate only ever applies from its start date forward. Changing a rate never re-prices usage that was already recorded under an earlier rate — usage before the change stays in its original segment at its original rate. That is why the timeline shows separate segments rather than re-costing the whole period at the latest rate.