← All posts

Guides

Google Translate API Pricing: What Should Buyers Budget?

Budget Google Translate API work by method, content volume and target languages. Compare the same translation deliverable with an enterprise model quote.

Google Translate API pricing needs a workload definition before it becomes an enterprise budget. Specify the Cloud Translation method, model, source volume, target languages and required output format. Then compare that purchase with a Token360 model quote using the same deliverable; a character price and a token price are not directly interchangeable.

A company translating product descriptions, help-center articles or supplier documents needs more than the lowest advertised unit rate. The quote must describe which material will be processed and what the application receives back. Otherwise, a cheap text result may be compared with a more complete document workflow.

This article is a documentation-based budget guide, not a translation benchmark. It covers the paid Cloud Translation API commonly searched as “Google Translate API,” rather than the consumer translation website. For the separate replacement decision, see whether an LLM can replace a translation API.

Which Google Translate API price applies to the job?

Choose the operation before choosing a rate. A text translation request, an adaptive translation request and a formatted-document job are different purchase specifications. The useful pricing comparison pairs a named method and model with its billing unit; the phrase “translation API” alone cannot tell a buyer which row of a price list to use.

Published reference, checked November 20, 2026: Google’s Cloud Translation pricing page lists standard NMT text at $20 per million input characters, with an eligible monthly $10 credit. Standard Translation LLM text is $10 per million input characters and $10 per million output characters. NMT document translation for the listed DOCX, PPT and PDF formats is $0.08 per page. Batch volume is multiplied by target languages. Method-specific eligibility and any contracted discounts still need checking.

Buyer’s jobSpecify before requesting a priceBudget mistake to avoid
Translate text fieldsModel, source volume and required language editionsProviding only the number of product records
Translate with examples or custom requirementsExact method and what context is submittedAssuming every translation method uses the same rate
Return translated documentsFile types, output format and layout acceptanceComparing extracted text with a finished document

Google’s text translation guide distinguishes Basic and Advanced capabilities and describes model selection. Procurement should retain the actual API configuration alongside the quote. A business user’s description of the task may be unchanged even when an engineering choice changes the service being purchased.

Gemini model-token rates are a separate budget topic. Our Google AI API pricing guide addresses that purchase; do not apply a Gemini price row to a Cloud Translation method simply because both are Google products.

How can a buyer turn a content inventory into a text estimate?

Build the estimate from the content actually scheduled for translation, rather than every record in the database. Separate the initial backlog from later updates, identify which languages each group needs, and document exclusions. This produces an auditable workload that finance can compare with usage records after a pilot instead of accepting an unexplained monthly allowance.

Illustrative calculation, not measured usage: suppose a batch contains 2 million source characters and requires three target languages. That represents 6 million character-language units. At the NMT reference rate above, gross processing is $120 before credits; if the full eligible $10 credit remains, the example becomes $110. Taxes, negotiated terms and other services are excluded. This is arithmetic, not a customer bill.

Keep the assumptions beside the result. If only one language needs an update, the application should not automatically describe the job as a complete three-language refresh. If every product description is rewritten each week, that is a different forecast from a one-time catalogue launch. A procurement spreadsheet should show the difference explicitly.

Use separate rows for:

  • Initial release: source material approved for the first language rollout.
  • Routine changes: new or changed content expected in an ordinary month.
  • Rework allowance: additional processing you choose to request after editorial review.
  • Optional expansion: extra languages or business units that are not yet committed.

These are planning categories, not claims about automatic caching, deduplication or free retries. Assign someone to explain why material is sent again. Without that ownership, the team may debate the supplier’s unit price while overlooking how its own publishing process creates additional work.

What changes when the company needs translated files back?

A document budget should name the returned artifact, not just the words inside the source file. If procurement expects a usable translated manual, specify the file type, layout requirements and review process. A text-only model response may cover the translation step while leaving extraction, formatting and delivery work for the application team.

Google’s Document Translation documentation describes direct translation of formatted files, including PDF and DOCX, with layout preservation. It also warns that scanned PDFs and complex layouts can lose formatting. Treat those limitations as reasons to test representative documents rather than as a promise that every uploaded manual returns ready to publish.

Choose a small acceptance set with the features your business actually uses: tables, multi-column pages, captions, product codes and scanned pages where applicable. Have the receiving team define what “usable” means before comparing suppliers. A result that needs manual reconstruction belongs in a different budget from one that can enter the publishing workflow after a normal language review.

For a Token360 model-based workflow, ask which parts your implementation must supply and which, if any, are included in the proposed engagement. Do not assume that purchasing language-model access includes document conversion or finished-file production. This guide has not tested either route or established a document-processing service in Token360.

How do you compare a character quote with a token quote?

Compare the cost of producing the same accepted deliverable, using actual usage from a representative pilot. Keep each supplier’s billing unit intact. A fixed characters-to-tokens conversion can hide differences in language, instructions, context and output length, so it should not determine which offer procurement approves for a multilingual production workload.

For a model-based translation pilot, record the complete submitted request, the measured input and output usage, and the returned material. Separate recurring instructions from the business content in your analysis, while retaining everything that was actually sent. Do not estimate usage from the source paragraph alone when the implementation adds a large terminology or formatting instruction.

The model must first satisfy the agreed business test. A low inference bill is not a saving if the resulting descriptions need unacceptable correction. Conversely, an existing translation service should not be rejected merely because another supplier uses a smaller-looking number with a different denominator.

Use a shared comparison sheet with the source set, requested language editions, required output format, measured API charges and review effort. Keep the review effort visible as an internal estimate; do not relabel it as a supplier fee. The general mechanics of a token quote remain in our LLM pricing comparison guide.

No pilot results are claimed here. The proposed evaluation can establish a comparable budget later; it cannot support a present claim that either route is cheaper or better.

What should an enterprise send Token360 for a translation quote?

Send a defined translation workload and ask Token360 to confirm model availability, usage assumptions and commercial scope. Include language pairs, expected volume, content type, required output and processing constraints. That lets the supplier price the model work while clearly identifying any application or editorial work outside the offer.

Information to provideWhat it lets the buyer verify
Language pairs and representative approved samplesThe proposed model can be evaluated on the actual task
Backlog, ordinary monthly changes and expected growthThe quote uses an agreed usage basis
Text response or finished-file requirementBoth suppliers are quoting the same deliverable
Instructions, terminology context and output constraintsThe pilot reflects the request the application will send
Data-handling and contractual requirementsUnconfirmed requirements remain visible before commitment
Acceptance owner and launch deadlineSomeone can approve quality and readiness against the brief

Token360 supports enterprise procurement across text, image, audio and video models. Enterprise pricing uses tiered quotes: higher usage receives lower quoted rates. Token360 does not charge a separate service fee. The actual quote must establish model coverage, eligible usage and terms; these statements do not imply free inference or guaranteed savings against Google.

Capacity, contracts and compliance requirements can be discussed for the engagement. Review the enterprise offering and API documentation before finalizing the integration scope. This article does not claim that Token360 resells Google Cloud Translation or supplies a drop-in equivalent of its document API.

What should be settled before approving the translation budget?

Approve a named service, a reproducible workload estimate and an acceptance plan together. The budget should distinguish committed usage from optional expansion and identify who authorizes changes. If the method, language coverage or required artifact changes, revisit the estimate instead of treating the old total as an enduring price promise.

Keep the dated price reference, quote, workload assumptions and pilot results in one approval record. An engineering owner should confirm the configuration; the content owner should confirm the output requirements; finance should confirm which usage is included in the forecast. Unknowns should remain explicit until the responsible person resolves them.

The goal is a purchase you can explain: this content, in these languages, through this service, producing this output under these terms. Once that scope is stable, a unit-price negotiation becomes meaningful. For company context, see About Token360.

Evidence and limits: Yuki Yu prepared this guide from linked official documentation and company-confirmed pricing positioning. The worked estimate is hypothetical. No quality, latency or savings test was performed. Formal publication URL and publication date remain pending.

  • Enterprise AI
  • API

Bring your next idea to life.

Create a video scene in Creator, or explore models for images, audio, and text to find the right capabilities for your project.

Get started