# Best Translation API for Indie Developers (2026)

> Best translation API for developers in 2026: DeepL, Google, Azure, Amazon and LLMs compared on price per 1M characters, free tiers, glossaries and data use.

- Canonical: https://zeroraised.com/blog/best-translation-api-for-indie-developers
- Published: 2026-11-13T13:30:00Z
- Author: ZeroRaised Team (https://zeroraised.com/blog/authors/zeroraised)
- Publisher: ZeroRaised (https://zeroraised.com/)
- Section: Building
- Tags: api, developer-tools, comparison, localization

**Short answer:** For classic machine translation at the lowest list price, use Azure AI Translator: $10 per 1M characters on S1, with 2M characters free each month on the F0 tier. Use DeepL if glossaries, formality control and clear data terms on a paid plan matter most. For translating user-generated content at runtime, a small LLM (Claude Haiku 5.5, gpt-5-nano or DeepSeek) is the cheapest per character in our example arithmetic, provided you cache every result.

Every price and limit below comes from the vendor's own pricing page or docs, checked on 13 Nov 2026. Anything we could not confirm on an official page today has been left out. Advice is ours and marked as such.

## TL;DR

- **Azure AI Translator:** $10 per 1M characters on S1 (East US), 2M free per month on F0 ([Azure pricing](https://azure.microsoft.com/en-us/pricing/details/translator/)).
- **Google Cloud Translation:** $20 per 1M characters for NMT after 500,000 free each month ([Google pricing](https://cloud.google.com/translate/pricing)).
- **Amazon Translate:** $15 per 1M characters, 2M free per month for 12 months ([AWS pricing](https://aws.amazon.com/translate/pricing/)).
- **DeepL API Free:** 500,000 characters per month; Pro is a fixed monthly price plus usage ([DeepL support](https://support.deepl.com/hc/en-us/articles/360021200939-DeepL-API-Free)).
- **LLMs** cost cents per 1M characters in our example arithmetic, but you own quality control and caching.
- **Self-host:** LibreTranslate (AGPL-3.0) on the Argos Translate library.

**Disclosure:** ZeroRaised publishes this guide and is a launch site for bootstrapped products. We have no commercial relationship with any vendor named here and earn nothing if you sign up. Facts come from the linked sources as checked on 13 Nov 2026.

## Which translation job do you actually have?

1. **UI strings at build time.** Small volume, translated once per release, quality matters most. Our guide on [how to localize a SaaS](https://zeroraised.com/blog/how-to-localize-a-saas) covers i18n setup, translation management tools and hreflang; this post is about the engine underneath.
2. **User-generated content at runtime.** Chat, reviews, support tickets. Volume grows with users and "good enough" is usually fine.
3. **Documents.** DOCX and PDF where layout must survive, often priced separately.

Our read: most indie apps start with job 1 and only hit real costs with job 2. Choose for the job that will grow.

## Translation APIs compared

| API | Free allowance | Entry paid price | Billing unit | Glossary / terminology | Data use (vendor's words, summarised) |
|---|---|---|---|---|---|
| Azure AI Translator | F0: 2M chars/month | S1: $10 per 1M chars (East US) | Characters | Custom Translator | Text translation not stored |
| Google Cloud Translation | 500K chars/month ($10 credit) | NMT: $20 per 1M chars | Characters (LLM: in and out) | Glossaries in Advanced (v3) | Not used for training |
| Amazon Translate | 2M chars/month for 12 months | $15 per 1M chars | Characters incl. whitespace | Custom terminology | May be used to improve; opt-out |
| DeepL API | Free: 500K chars/month | Pro: monthly fee plus usage | Characters | Glossaries; formality (some languages) | Paid plans: not stored or trained on |
| LLM APIs | Varies (Gemini has a free tier) | From $0.05 per 1M input tokens (gpt-5-nano) | Tokens in and out | Put the glossary in the prompt | Varies by provider and tier |
| LibreTranslate (self-host) | Free software | Your server costs | n/a | n/a | Stays on your server |

Each row is sourced in the vendor sections below. The language pages we checked do not state a single total count, so check your exact pairs on [Azure's list](https://learn.microsoft.com/en-us/azure/ai-services/translator/language-support), [Amazon's list](https://docs.aws.amazon.com/translate/latest/dg/what-is-languages.html) and [DeepL's list](https://developers.deepl.com/docs/getting-started/supported-languages) (DeepL's `/v3/languages` endpoint also reports per-language feature support).

## DeepL API

DeepL API Free lets you translate up to 500,000 characters a month and stops at that cap. API Pro has no free characters, no volume cap, and bills a fixed monthly price plus usage per translated character ([DeepL support](https://support.deepl.com/hc/en-us/articles/360021200939-DeepL-API-Free)).

We could not load DeepL's API Pro per-character rate from an official DeepL page today, and we do not copy prices from third-party sites, so check [DeepL's pricing page](https://www.deepl.com/en/pro) (API tab) in your currency. What you get:

- **Glossaries** to pin brand names and product terms.
- **Formality** control, which DeepL says is "only available for some target languages" ([request translation reference](https://developers.deepl.com/api-reference/translate/request-translation)).
- **Cost control:** you can set a monthly character limit on a Pro subscription to cap spend ([DeepL cost control](https://developers.deepl.com/docs/best-practices/cost-control)).
- **Data handling:** DeepL's [API overview](https://developers.deepl.com/docs/getting-started/about) says "With DeepL API paid plans, texts aren't saved on persistent storage and aren't used to train our models." Its privacy page adds that "Data handling may differ between the Free and Pro plans" ([DeepL privacy for developers](https://developers.deepl.com/docs/resources/privacy)).

Our read: if you translate anything private (support tickets, messages), do not run it through DeepL Free. Read DeepL's terms for Free yourself, or start on Pro.

## Google Cloud Translation

Google's [pricing page](https://cloud.google.com/translate/pricing) lists NMT text translation (Basic and Advanced) at $20 per 1M characters above the first 500,000 characters each month, which are free and "applied as $10 credit every month".

Its Translation LLM costs $10 per 1M input characters plus $10 per 1M output characters, which Google calls "cost equivalent with NMT". Custom (AutoML) models start at $80 per 1M characters, and empty requests are charged as one character, so guard against blank strings.

Glossaries live in Cloud Translation Advanced (the v3 API) ([Google glossary docs](https://docs.cloud.google.com/translate/docs/advanced/glossary)). On data, Google says: "Google does not use the content you send to train and improve our Google Translation features," and that text "is held briefly in-memory in order to perform the translation" ([Google data usage](https://docs.cloud.google.com/translate/data-usage)).

## Azure AI Translator

Azure's [pricing page](https://azure.microsoft.com/en-us/pricing/details/translator/) describes the F0 tier as "2 million characters of any combination of standard translation and custom translation training free per month". The page did not render S1 dollar amounts for us, so we took them from Microsoft's public [Azure Retail Prices API](https://prices.azure.com/api/retail/prices) (East US region): standard text translation is $10 per 1M characters, document translation $15, and custom text translation $40, plus $10 per month to host a custom model. Prices can differ by region. On data, Microsoft says "Text translation processes customer data at REST and doesn't store customer data," and that document translation stores files only while processing and then hard-deletes them ([Azure data privacy](https://learn.microsoft.com/en-us/azure/ai-foundry/responsible-ai/translator/data-privacy-security)).

Our read: Azure is half of Google's NMT list rate and F0 is the most generous recurring free allowance here. It is the default we would test first.

## Amazon Translate

Amazon's [pricing page](https://aws.amazon.com/translate/pricing/) lists real-time text translation at **$15 per 1M characters**, with 2M characters per month free for 12 months. Batch document translation is also $15, real-time DOCX is $30 and Active Custom Translation is $60 per 1M characters. Amazon may charge for automatic language detection ([Amazon language docs](https://docs.aws.amazon.com/translate/latest/dg/what-is-languages.html)). **Custom terminology** pins brand and product names, though Amazon "doesn't guarantee that it will use the target term for every translation" ([custom terminology docs](https://docs.aws.amazon.com/translate/latest/dg/how-custom-terminology.html)).

The data terms are the main difference. Amazon's [FAQ](https://aws.amazon.com/translate/faqs/) says it "may store and use text inputs processed by the service", including to "improve and develop the quality of Amazon Translate and other Amazon machine-learning/artificial-intelligence technologies". You can opt out "by using an AWS Organizations opt-out policy." Set that up before you send user content.

## LLM APIs as translators

LLMs are priced per token, not per character. List prices per 1M tokens:

| Model | Input | Output | Source |
|---|---|---|---|
| OpenAI gpt-5-nano | $0.05 | $0.40 | [OpenAI](https://developers.openai.com/api/docs/pricing) |
| OpenAI gpt-5-mini | $0.25 | $2.00 | [OpenAI](https://developers.openai.com/api/docs/pricing) |
| Claude Haiku 5.5 (prompts up to 100K tokens) | $0.10 | $0.50 | [Anthropic](https://platform.claude.com/docs/en/about-claude/pricing) |
| Claude Sonnet 5.5 | $2 | $10 | [Anthropic](https://platform.claude.com/docs/en/about-claude/pricing) |
| Gemini 3.1 Flash-Lite (paid tier) | $0.25 | $1.50 | [Google AI](https://aistudio.google.com/docs/pricing) |
| DeepSeek deepseek-flash (off-peak, cache miss) | $0.15 | $0.60 | [DeepSeek](https://api-docs.deepseek.com/quick_start/pricing) |

OpenAI and Anthropic list batch processing at 50% off. DeepSeek charges double in weekday peak hours (01:00 to 04:00 and 06:00 to 10:00 UTC). Google's pricing page says Gemini free-tier data is used to improve its products; paid-tier data is not. Our post on [whether AI APIs train on your data](https://zeroraised.com/blog/do-ai-apis-train-on-your-data) goes through each provider's terms.

**Example arithmetic, not a quote.** Assume 1M characters of English source is about 250,000 input tokens (roughly 4 characters per token) and the translation comes back as about 300,000 output tokens, ignoring prompt overhead. Token ratios vary a lot by language and script, so measure your own.

On those assumptions, 1M source characters costs about $0.13 on gpt-5-nano (0.25 x $0.05 + 0.3 x $0.40), $0.18 on Claude Haiku 5.5, $0.22 on DeepSeek flash off-peak, $0.51 on Gemini 3.1 Flash-Lite, $0.66 on gpt-5-mini and $3.50 on Claude Sonnet 5.5. Compare $10 on Azure S1, $15 on Amazon and $20 on Google NMT.

Our read: the cost gap is real, but the LLM route needs more engineering. Pass glossary terms in the prompt, tell the model to keep placeholders like `{name}` and HTML tags intact, and check every placeholder survives. If LLM spend starts to bite, our guide to [reducing LLM costs](https://zeroraised.com/blog/how-to-reduce-llm-costs) covers batching, caching and model routing.

## Self-hosting: LibreTranslate and Argos

[LibreTranslate](https://github.com/LibreTranslate/LibreTranslate) is an open source translation API under the GNU AGPL v3, and its engine "is powered by the open source Argos Translate library". [Argos Translate](https://github.com/argosopentech/argos-translate) itself is "dual licensed under either the MIT License or Creative Commons CC0" and runs offline using OpenNMT models.

Our read: self-host when text must never leave your servers. Read the AGPL before modifying LibreTranslate and offering it over a network; embedding Argos as a library avoids that question.

## Caching: translate once, store forever

The cheapest translation is the one you never repeat. A simple pattern works for every API above:

```ts
// key = hash of everything that changes the output
const key = sha256([engine, engineVersion, glossaryVersion, sourceLang, targetLang, text].join("|"));

const cached = await db.translations.find(key);
if (cached) return cached.text;

if (!text.trim() || sourceLang === targetLang) return text; // skip empty and same-language calls

const translated = await translate(text, sourceLang, targetLang);
await db.translations.insert({ key, text: translated });
return translated;
```

Rules we would follow:

1. **UI strings:** translate in CI and only send new or edited keys.
2. **User content:** translate lazily, the first time a reader in another language opens it.
3. **Detect first:** skip messages already in the reader's language (Amazon may bill auto-detection).
4. **Version the key** so an engine or glossary change retranslates gradually.
5. **Cap spend** with DeepL's monthly character limit or a cloud budget alert.

## Worked example: a small community app

**Example, not a quote.** Your app translates 3M characters a month of reviews and tickets into English, after caching.

| Option | Monthly cost on list prices |
|---|---|
| Google NMT | (3M minus 500K free) x $20 per 1M = $50 |
| Amazon Translate, first 12 months | (3M minus 2M free) x $15 per 1M = $15 |
| Amazon Translate, after 12 months | 3M x $15 per 1M = $45 |
| Azure S1 | 3M x $10 per 1M = $30 |
| Azure F0 | Free, but only up to 2M characters a month, so not enough here |
| Claude Haiku 5.5 (our token assumptions above) | 3 x $0.18 = about $0.54, plus prompt overhead |
| gpt-5-nano (same assumptions) | 3 x $0.13 = about $0.39, plus prompt overhead |

At 1.5M characters a month, Azure F0 would cover the whole bill. Our read: start on a free tier, add caching on day one, and only move to an LLM if the classic API bill becomes a meaningful line item.

## Decision table

| Your situation | Our pick |
|---|---|
| Under 2M characters a month, want $0 | Azure F0 |
| Lowest list price for classic machine translation | Azure S1 ($10 per 1M) |
| Need glossaries and formality, care about paid-plan data terms | DeepL API Pro |
| Already on Google Cloud | Google Cloud Translation Advanced |
| Already on AWS, in your first year | Amazon Translate, with the AI opt-out policy set |
| High-volume chat or reviews, cost-sensitive | Small LLM (Haiku 5.5, gpt-5-nano, DeepSeek) plus caching |
| Text must never leave your servers | Self-host LibreTranslate or Argos |
| Marketing pages and onboarding copy | Any engine plus human review |

## Frequently asked questions

### What is the best translation API for indie developers?

For classic machine translation at the lowest list price, Azure AI Translator: $10 per 1M characters on S1 and 2M characters free per month on F0. Pick DeepL if you need glossaries and formality control and want paid-plan texts kept out of storage and training. For user-generated content at runtime, a small LLM such as Claude Haiku 5.5 or gpt-5-nano is cheapest per character, as long as you cache every translation.

### Is there a free translation API?

Several have free allowances. Azure F0 gives 2M characters per month, DeepL API Free gives 500,000 characters per month, Google Cloud Translation gives the first 500,000 characters each month as a $10 credit, and Amazon Translate gives 2M characters per month for the first 12 months. LibreTranslate is open source (AGPL-3.0) and free to self-host.

### Is DeepL better than Google Translate API?

It depends on your languages and content, so test both on your own strings. On price, Google NMT is $20 per 1M characters after 500,000 free. DeepL Pro is billed as a fixed monthly price plus a per-character usage charge; check DeepL's API pricing page for the current rate in your currency. DeepL offers glossaries and a formality setting for some target languages; Google offers glossaries on Cloud Translation Advanced.

### Is it cheaper to translate with an LLM like GPT or Claude?

Usually, on list price. Using our stated assumption of about 4 characters per token, 1M characters costs roughly $0.13 with gpt-5-nano or $0.18 with Claude Haiku 5.5, versus $10 to $20 on classic APIs. The trade-offs are prompt overhead, less predictable output, no built-in glossary feature, and the need to check each provider's data terms.

### Do translation APIs keep or train on my users' text?

Terms differ. Google says it does not use content sent to Cloud Translation to train its translation features. Microsoft says Azure text translation does not store customer data. DeepL says paid API plans do not save texts on persistent storage or use them for training, and warns Free and Pro data handling may differ. Amazon may store and use inputs to improve the service unless you opt out with an AWS Organizations policy.

## Launch in every language

Shipping a multilingual app? Run through our [product launch checklist for bootstrapped makers](https://zeroraised.com/blog/product-launch-checklist-for-bootstrapped-makers), then [submit your product to ZeroRaised](https://zeroraised.com/submit) for free. For more developer tool comparisons, [join the ZeroRaised newsletter](https://zeroraised.com/newsletter).

## Sources

All last checked 13 Nov 2026.

- [DeepL API: about the API](https://developers.deepl.com/docs/getting-started/about)
- [DeepL API Free (DeepL support)](https://support.deepl.com/hc/en-us/articles/360021200939-DeepL-API-Free)
- [DeepL developer privacy](https://developers.deepl.com/docs/resources/privacy)
- [DeepL subscription-level cost control](https://developers.deepl.com/docs/best-practices/cost-control)
- [DeepL request translation reference](https://developers.deepl.com/api-reference/translate/request-translation)
- [DeepL supported languages](https://developers.deepl.com/docs/getting-started/supported-languages)
- [DeepL pricing](https://www.deepl.com/en/pro)
- [Google Cloud Translation pricing](https://cloud.google.com/translate/pricing)
- [Google Cloud Translation glossaries](https://docs.cloud.google.com/translate/docs/advanced/glossary)
- [Google Cloud Translation data usage](https://docs.cloud.google.com/translate/data-usage)
- [Azure AI Translator pricing](https://azure.microsoft.com/en-us/pricing/details/translator/)
- [Azure Retail Prices API](https://prices.azure.com/api/retail/prices)
- [Azure Translator language support](https://learn.microsoft.com/en-us/azure/ai-services/translator/language-support)
- [Azure Translator data, privacy and security](https://learn.microsoft.com/en-us/azure/ai-foundry/responsible-ai/translator/data-privacy-security)
- [Amazon Translate pricing](https://aws.amazon.com/translate/pricing/)
- [Amazon Translate FAQ](https://aws.amazon.com/translate/faqs/)
- [Amazon Translate supported languages](https://docs.aws.amazon.com/translate/latest/dg/what-is-languages.html)
- [Amazon Translate custom terminology](https://docs.aws.amazon.com/translate/latest/dg/how-custom-terminology.html)
- [OpenAI API pricing](https://developers.openai.com/api/docs/pricing)
- [Anthropic Claude pricing](https://platform.claude.com/docs/en/about-claude/pricing)
- [Gemini API pricing](https://aistudio.google.com/docs/pricing)
- [DeepSeek API pricing](https://api-docs.deepseek.com/quick_start/pricing)
- [LibreTranslate on GitHub](https://github.com/LibreTranslate/LibreTranslate)
- [Argos Translate on GitHub](https://github.com/argosopentech/argos-translate)

## Frequently asked questions

### What is the best translation API for indie developers?

For classic machine translation at the lowest list price, Azure AI Translator: $10 per 1M characters on S1 and 2M characters free per month on F0. Pick DeepL if you need glossaries and formality control and want paid-plan texts kept out of storage and training. For user-generated content at runtime, a small LLM such as Claude Haiku 5.5 or gpt-5-nano is cheapest per character, as long as you cache every translation.

### Is there a free translation API?

Several have free allowances. Azure F0 gives 2M characters per month, DeepL API Free gives 500,000 characters per month, Google Cloud Translation gives the first 500,000 characters each month as a $10 credit, and Amazon Translate gives 2M characters per month for the first 12 months. LibreTranslate is open source (AGPL-3.0) and free to self-host.

### Is DeepL better than Google Translate API?

It depends on your languages and content, so test both on your own strings. On price, Google NMT is $20 per 1M characters after 500,000 free. DeepL Pro is billed as a fixed monthly price plus a per-character usage charge; check DeepL's API pricing page for the current rate in your currency. DeepL offers glossaries and a formality setting for some target languages; Google offers glossaries on Cloud Translation Advanced.

### Is it cheaper to translate with an LLM like GPT or Claude?

Usually, on list price. Using our stated assumption of about 4 characters per token, 1M characters costs roughly $0.13 with gpt-5-nano or $0.18 with Claude Haiku 5.5, versus $10 to $20 on classic APIs. The trade-offs are prompt overhead, less predictable output, no built-in glossary feature, and the need to check each provider's data terms.

### Do translation APIs keep or train on my users' text?

Terms differ. Google says it does not use content sent to Cloud Translation to train its translation features. Microsoft says Azure text translation does not store customer data. DeepL says paid API plans do not save texts on persistent storage or use them for training, and warns Free and Pro data handling may differ. Amazon may store and use inputs to improve the service unless you opt out with an AWS Organizations policy.