Get API key
PlatformTourPricingEnterpriseDocsSign in

Pay per job

Characters + model usage at cost

$2.50 / 1,000 characters

+ model usage at cost, no markup

  • Failed jobs are free
  • Memory reused word for word isn't charged
  • Price fixed on delivery
  • Bring your own model provider key
Get API key

Estimate a month

Already in your translation memory, reused word for word
Characters, over every target language
At $2.50 / 1,000
From your memory
Model usage, on average
A month
About these numbers

Model usage is at the average jobs have cost so far, about $0.006 per 1,000 characters; a job's own runs from $0.001 to $0.016, with the models you choose, the languages and the size of the job. It stays as it is for memory reused: the model still reads a reused string and decides to keep it. How much of your text your memory already holds depends on what you send: on the jobs we have delivered, from under 1% for new content to 64% for a project that sends its whole string file again with each update.

For scale, a year of orders by localization clients, as a month on average, characters summed over every target language: the top tenth 207,000, the top quarter 49,700, the mean client 100,100, the median 10,300, the bottom quarter 2,700, the bottom tenth 600. Orders are irregular, so a month with work is larger than the average one: the median client orders in two months of the year, about 75,000 characters in each — $187.50 and $0.44 of model usage — and the average client about 290,000 in a month with work: $725.00 and $1.72.

How it is charged

Characters

The text of every string you send, once per target language: 1,000 characters into three languages is 3,000. Keys, context notes and screenshots aren't counted as characters; the model work they take is in model usage.

Your memory, reused, isn't charged

A string delivered word for word as a translation your memory already holds for it is not charged the $2.50 rate: you paid for that translation once. It is checked against your memory, not taken on the model's word. The model still read it and decided to keep it, so its model usage is charged at cost like any call. Each job shows the split: characters charged, and characters from your memory.

Model usage, one to one

Every call your job makes — translating, choosing the terms and screens each string is shown, checking your terminology, searching your memory — at exactly what the provider charged for it. Each call is listed with its cost under Model calls, so the figure can be checked, not taken on trust.

Or on your own key

Give us your own model provider key and every call is made on your account, billed to you by your provider directly — your price is then characters only. Your cabinet shows what went through it.

Fixed when delivered

The price is set as the job is delivered, the total rounded to the cent. A later change to your plan doesn't reprice work you already have.

Failed jobs aren't charged

Neither their characters nor their model usage, and neither is a job still in progress.

Visible per job, and by month

Your cabinet shows each job's characters, those from your memory, model usage and price, and the Usage tab adds them up by month, by project and by API key; GET /v1/jobs returns them as characters, tm_characters, billing.characters_usd, billing.models_usd and price_usd.

White label, on-premises, your own models Enterprise →