The right model for every language, and every kind of text.
For teams that ship a product in many languages. Each language, and each kind of text within it, goes to the model you choose, with your glossary, translation memory, style guide, screenshots and anything else you keep beside every string, and comes back checked, in the format it was sent in.
- Glossary
- Translation memory
- Style guide
- Screenshots
- Design specs
- Product docs
- Reviewer notes
- Locale rules
- Reference sites
- Support macros
es-ESuiObtener clave de APIGemini 3.8 Flashes-ESmarketingTu trabajo, en todos los idiomas.Mistral Large 4it-ITuiOttieni chiave APIGPT-5.6 Lunait-ITmarketingIl tuo lavoro, in ogni lingua.Claude Opus 5.5ar-SAuiالحصول على مفتاح واجهة برمجة التطبيقاتGrok 4.7ko-KRuiAPI 키 가져오기Qwen3.7 Flashhi-INuiएपीआई पासकोड पानाGemini 4 ArgonWhich model translates best, by language and kind of text?
MQM penalty per 1,000 words of English source, scored by professional linguists.
AI in a typical TMS vs Localize.io
| What | A typical TMS with AI | Localize.io |
|---|---|---|
| What it costs | A typical TMSModel spend is often hidden, or shown only as a total. | Localize.ioEvery model call is recorded with its cost. See it in the tour: The record |
| Long files | A typical TMSTypically sent in one request: a long answer gets cut off, and is still paid for. | Localize.ioSplit into batches sized to each model's output limit. See it in the tour: Send work |
| When a check fails | A typical TMSThe model is often re-asked silently, and billed again. | Localize.ioEvery attempt is logged, at most three; a string that still fails comes back flagged, and isn't charged the character rate. See it in the tour: Checks |
| What the model did | A typical TMSPrompts, reasoning and tokens typically stay inside the platform. | Localize.ioThe prompt, the reply, the tokens and the cost are visible on every call. See it in the tour: The record |
| Which model | A typical TMSThe platform picks the model, and how it decides isn't shown. | Localize.ioYour configuration picks the model, per language, content type, project and pass, and every job shows which setting decided it. See it in the tour: Routing |
Routing
Different work goes to different models.
A button label, a help article and a campaign headline are different jobs, even in the same language. So you choose the model per language, per content type, per project and per pass: a quick, inexpensive one for short interface strings, a stronger one where the wording has to sell, and a second one to review a draft if you want one.
You set it once in your configuration. Change it whenever you like: the next job follows the new route, and nothing else has to be rebuilt.
A route is set by
ui and Spanish marketing apart.es-ES marketingMistral Large 4GPT-5.6 Solfrom the next job
The whole job
Hand over what you have. Get back what you sent.
Picking a model is the small part. Someone also has to convert files, sort the reference material, split the work, choose what goes with each string, check what comes back and redo what fails. Here an agent does all of it.
Upload whatever context you keep, as it is, in your own folders and formats: glossaries, translation memories, style guides, brand books, screenshots, notes for translators. Then send strings or a file. The agent works out what each asset is, prepares and batches the work, gathers the context for each batch, sends it to the model its route names, has the result checked, and returns it in the format it arrived in. Your files are never changed.
You hand over
Terminology 2025.xlsx
Writing guidelines.pdf
past/help-centre_es.xliff
screens/plans.png
notes for translators.md
You send, and get back
strings/en.jsonwhat you sendstrings/es.jsonthe same keys, the same shapeContext
Context where it helps, and nowhere else.
Nothing is attached by default. For each batch, the agent decides what each string needs: a screenshot where the screen explains it, a glossary term only for a string that is about that term, a past translation only where one is close in meaning. A short label, a support answer and a marketing line each get the context that suits them.
Your memory is a guide, not an answer key. Wording you approved carries forward; an entry that is out of date or wrong isn't repeated.
Step 2: your assetsThe examplesExample 3: translation memory
Three strings, three different contexts
A button or a noun? The screen says: the button under the Pro plan.
The name of the page has to match the name in your product.
A close line from last year's campaign: its tone carries over, “business” doesn't.
- Less setupNo converting, sorting or preparing first. Bring what you have.
- New material counts from the next jobAdd a glossary or screenshots for the whole account or one project, whenever you like.
- Every result is checkedBefore it comes back to you, not after someone notices.
- A full recordHow each translation was made: the route, what the model was shown, what it answered, the checks, the cost.
- Fits how you workA REST API for your pipeline, MCP for your agents.
Connect in minutes. One key per integration, over REST or MCP.
Read the docsTranslate your first strings in a few minutes.
White label, on-premises or your own models? Enterprise →