Skip to content
Finitizer

AI Usage Export Analyzer

See your AI spend by model, project and day, the spikes, and what batch or a cheaper model would save.

AI & TokensReads your own dataBring a file
New here? See what this tool does and how to use it.

Learn more

AI Usage Export Analyzer

What this free tool does for you

See where your AI money really goes: which model, project or API key spent it, and which days spiked. Then see what moving work that can wait to half-price batch, or to a cheaper model, would save.

How to use it

  1. Export a usage or cost CSV from the OpenAI usage dashboard or the Claude Console. No file yet? Click Load a worked example.
  2. Click Choose usage exports, or drop the files in. Load both vendors together to see all your AI spend.
  3. On each file, open How the columns were read and fix any column that was read wrong.
  4. Read the AI spend in these files, then Spend by model and Spend by day for the days that spiked.
  5. Under What would change the bill, see the Saving from batch and what moving to another model would save.

Good to know

  • Private. Your files are read in this browser tab and never sent anywhere.
  • Will it match the invoice? Not exactly. Taxes, credits and commitments are not in usage exports, so use the cost export when you need the billed figure.
  • On OpenAI Enterprise? Since April 1, 2026, Enterprise invoices no longer itemize API costs, so the usage export is where that detail lives.
Before you start: what to have readyShow
  • A usage or cost export as CSV: from the OpenAI usage dashboard, the Claude Console, or a gateway that logs model and tokens.
  • Ideally a whole month at a one-day interval, so the daily figures and the unusual-day check have something to compare.
  • Both vendors at once if you use both: load the files together.

Or drop files here: the OpenAI usage dashboard’s cost or activity export, the Claude Console’s cost or usage export, or any CSV with a model and a cost or token column. Load several at once to see every vendor together.

Your files are read in this browser tab and never sent anywhere.

Prices verified 12 Oct 2026. Public list prices in USD, before any discount you have negotiated.

Paying negotiated rates for Claude, GPT or Gemini?

Cloud and AI Invoice Reconciliation: We check every AWS, Google Cloud and AI invoice against your usage and contracts each month. From $1,250 a month. Nothing paid upfront.

See the service

About this analyzer

The calculators on this site price a workload you describe. This reads the one you actually ran: which models the money went to, which project or key spent it, which day it jumped, how much of your input already comes from cache, and how much of your spend goes through the batch API, which is half price on most models.

The two what-ifs use the same verified list prices as the LLM Token Cost Calculator: moving the traffic that can wait to the batch API, and running one model’s traffic on another, token for token.

Questions

Which files does it read?

The OpenAI usage dashboard’s exports (cost data grouped by line item, and activity data grouped by project, user, API key, model or batch), the Claude Console’s cost and usage exports, and any CSV with a model column and a cost or token column, such as a gateway or proxy log. Load several at once to see every vendor in one view. Columns are matched by name, and each file shows what every column was read as, so you can correct it.

How do I export my OpenAI usage?

In the OpenAI platform, open Usage and choose Export. For spend that reconciles to the invoice, use the Cost data tab, group by line item, all projects, CSV, and a one-day interval. For tokens by project, API key or model, use the Activity data tab and group by the one you want. Since April 1, 2026, Enterprise invoices no longer itemize API costs, so this export is where that detail lives.

How do I export my Claude usage?

In the Claude Console, the Usage page exports tokens for the filters on screen (workspace, model, month, API key) and the Cost page exports spend by model and day, web search and code execution included. Set the filters first: the file holds what the page shows. You need the Developer, Billing or Admin role.

Why are some figures marked as list price?

An activity export has token counts and no cost. For those rows the tool prices the tokens at the public list rates in our LLM price index, before any discount you have negotiated. A cost a file states is used as it is and never re-priced, and the result says how much of the total came from each.

Why does the total not match my invoice?

Taxes, credits and commitments are not in usage exports, and month boundaries are in UTC. Rows priced at list price also miss any negotiated discount and the long-context surcharge some models add above a prompt length. Use the cost export when you need the billed figure.

Is my file uploaded anywhere?

No. The files are read by your browser and analyzed in the tab; nothing in them is sent. Usage statistics record that a file was loaded, its detected type and a size band, never its name or contents. The /privacy page lists every field.

What makes a day unusual?

A day whose spend is more than twice the median of the seven days before it, and higher than that median by at least half an average day, so a few cents on a quiet account is not flagged. Days with no usage count as zero. The rule needs more than a week of data.

Related tools

More tools for: AI and model usage, Billing data, allocation and governance

One export is a snapshot. AI spend moves every day.

Finitizer Tokenomics tracks token spend by model, team and feature as it happens, alongside your cloud bill, and flags the runaway loop on the day it starts rather than at month end.

See how Finitizer works