Translation API Pricing 2026: DeepL vs Google Cloud vs Azure per Million Characters¶

You build the integration, test it on sample data, calculate your projected costs - and then the first real invoice arrives at three times what you budgeted. The reason is usually simple: you counted content characters, but the API counted everything - HTML tags, whitespace, metadata. And DeepL has a per-document minimum charge that’s mentioned quietly in the documentation.

Here’s the full breakdown before those surprises hit your billing dashboard.

All Three Providers in One Table¶

Starting with the simplest view - what each provider charges per million characters under standard usage:

This is just the starting point. Actual costs depend on content type, volume, whether you need custom models, and whether you already have a Microsoft Azure or Google Cloud account.

Google Cloud Translation: $20/M and 249 Languages¶

Google Cloud Translation API comes in two versions: Basic (v2) and Advanced (v3). For standard NMT (Neural Machine Translation) the pricing is identical: $20 per million characters after the first 500K free per month.

What’s included at $20/M:

• 249 languages - widest coverage of the three providers
• Automatic language detection
• Batch translation
• Glossaries - custom term dictionaries for brand consistency

Cloud Translation Advanced adds premium options:

• LLM-based Adaptive Translation: $25/M input + $25/M output - so $50/M in practice
• Custom AutoML models: from $80/M at volumes up to 250M characters/month
• Document translation (Word, PPT, PDF): $0.08 per page via the Document Translation API

Volume discounts kick in above 1 billion characters per month - but that means a conversation with Google’s sales team.

The character counting gotcha:

Google bills all characters including whitespace, newlines, and HTML tags when you send HTML. No discounts for “empty” characters. For markup-heavy content, your actual bill can run 40-100% higher than the visible text character count.

Where Google has a clear edge is language coverage. Japanese, Chinese, Korean, Arabic, most African and Southeast Asian languages - all covered, with decent quality. For multilingual products where European languages are just part of the picture, there’s really no substitute for Google’s breadth.

Azure Translator: $10/M - the Cheapest Option¶

Azure Translator is the cheapest of the three by base price: $10 per million characters at the standard S1 tier. It also has the most generous free tier: 2 million characters per month - but only for your first year of Azure.

Tier breakdown:

What’s included at $10/M:

• 130+ languages
• Transliteration (script conversion - e.g. Cyrillic to Latin)
• Language detection
• Dictionary lookup
• Document translation (async batch) via a separate endpoint

Where Azure clearly wins:

If your company is already on Microsoft Azure, integration is straightforward - Logic Apps, Power Automate, Azure Functions, Teams all connect natively. For Microsoft-stack shops it’s the natural choice with unified billing.

Where Azure falls short:

Translation quality for European language pairs is noticeably below DeepL. For high-visibility content - UI copy, marketing, client-facing legal documents - the difference will be visible to end readers. For internal operations, structured technical text, and bulk content the gap matters less.

Critical note: the “2M characters free” only applies to year one. After that, billing starts from the first character at the full $10/M rate. If you modeled your budget based on year-one pricing and forgot to recalculate, that’s a jarring month-13 invoice.

DeepL API: More Expensive, but for a Reason¶

DeepL is the most expensive of the three, but independent benchmarks consistently rate it highest for European language pairs. In 2026, DeepL supports 33 languages - far fewer than Google or Azure, but within that set (EN, DE, FR, ES, IT, NL, PL, UK and their cross-pairs) it delivers the most natural output.

DeepL offers several API tiers:

Developer Plan (free)¶

• 500K characters per month
• Rate-limited
• Good for development and testing

Growth Plan ($26/month)¶

• 1M characters/month included
• Overage: $27.50 per additional million
• Up to 10 API keys, 2,000 glossaries
• CAT tool integration (SDL Trados, memoQ, Phrase)
• Speech-to-text: 10 hours/month included

API Pro (pay-as-you-go)¶

• $5.49/month base fee
• $25 per million characters above the 500K free tier
• No cap on volume

Cost comparison at 5M characters/month:

• Growth: $26 + (4M × $27.50) = $136/month
• API Pro: $5.49 + (4.5M × $25) = $117.99/month

At 10M characters/month:

• Growth: $26 + (9M × $27.50) = $273.50/month
• API Pro: $5.49 + (9.5M × $25) = $243.49/month

Above roughly 3-4M characters/month, API Pro is cheaper than Growth because the overage rate is lower ($25 vs $27.50). But Growth is more team-friendly with more API keys and CAT integration.

Quality in numbers:

As the Intlpull 2026 machine translation benchmark reports:

DeepL led in 65% of language pairs tested, with 10 errors versus Google’s 25 in professional evaluations for European language pairs.

A developer on a localization forum shared their A/B test:

I ran all three APIs on the same 1000 Ukrainian-to-German sentences. DeepL won 7 out of 10 human evaluations, Google 2, Azure 1. For Japanese-to-English though, Google won easily.

If your core languages are European, DeepL justifies the premium. If Asian and European languages are mixed, a hybrid setup (DeepL for European, Google for Asian) gives the best quality-to-cost ratio - but it requires more infrastructure.

The document translation gotcha:

When translating files (Word, PowerPoint, PDF) via the DeepL Document Translation API, a minimum of 50,000 characters per document applies regardless of actual content size. A 2,000-character Word doc costs the same as a 50,000-character one.

Plain text formats (HTML, TXT, SRT, XLIFF) have no minimum - you pay for actual character count.

Hidden Costs: What Actually Shows Up on the Invoice¶

Base prices only tell part of the story. Here’s what actually drives your final bill.

HTML and XML Tags¶

Google Cloud Translation: bills all characters including HTML tags. <p class="intro"> is already 18 billable characters. For web content or CMS-managed text with rich markup, real bills run 1.5-2x the visible text character count.

Azure Translator: same - tags are counted.

DeepL API: with tag_handling=html or tag_handling=xml, DeepL ignores characters inside tags and counts only text content. For HTML-heavy content this is a real cost advantage.

As BuildMVPFast notes in their API comparison:

Google, Azure, and Amazon count HTML tags and attributes as characters. For HTML-heavy content, this can double or triple your character count. DeepL strips XML/HTML tags before counting, which can be a significant cost advantage.

Best practice: always send pure text when possible. If not - use tag_handling in DeepL, and strip HTML before sending to Google or Azure. For typical CMS content, this reduces your bill by 20-40%.

Document Minimum at DeepL¶

Worth emphasizing: 50K character minimum per Word/PPT/PDF file in DeepL. If you’re processing many small documents, this is a hidden cost multiplier. For SaaS products where users upload short CVs or single-page contracts, actual costs can run 10-25x the per-character estimate.

Fix: convert documents to HTML or plain text before sending to DeepL - that way actual character count applies.

Whitespace¶

Spaces, tabs, newlines - all counted by all three providers. If you’re generating pretty-printed JSON or indented HTML before sending, you’re paying for the indentation.

Fix: minimize whitespace before the API call. In Python: json.dumps(data, separators=(',', ':')) instead of default formatting. In practice this saves 5-15% depending on data structure.

LLM Modes Cost More¶

Google Cloud offers LLM-based Adaptive Translation at $25/M input + $25/M output - so effectively $50/M, which is 2.5x the base NMT rate. DeepL’s next-gen model is included in the standard plan pricing.

Real Costs at Different Volumes¶

Monthly costs for common usage scenarios (post-year-one, no Azure free tier):

What stands out:

• Under 500K/month: all effectively free (Azure gives 2M/month year one). DeepL Growth is expensive relative to usage.
• 1M to 10M/month: Azure is half the cost of Google, and 2.5-3x cheaper than DeepL.
• 50M+/month: Azure and Google converge with commitment discounts; DeepL requires Enterprise negotiations.

As SimpleLocalize summarizes:

For 10,000 translation keys across 5 languages at ~60 characters per key: Google NMT ~$60, DeepL API ~$88, Azure ~$50. The gap widens significantly at higher volumes.

How to Choose: When Google, When Azure, When DeepL¶

There’s no single right answer - it depends on your specific use case.

Choose Google Cloud Translation if:¶

• You need languages DeepL doesn’t cover: Japanese, Chinese, Korean, Arabic, most African and Southeast Asian languages
• You’re building a multilingual product across 10+ languages where most aren’t European
• You need Custom AutoML models for a specific domain (legal, medical, technical)
• You’re already on Google Cloud and want unified billing

Choose Azure Translator if:¶

• You’re already on Microsoft Azure and want a single ecosystem bill
• Cost is the primary driver and quality is secondary
• Volumes are 10M+ characters/month where $10/M makes a meaningful difference
• You need native integration with Microsoft 365, Teams, or Power Automate

Choose DeepL API if:¶

• Your core languages are European (en, de, fr, es, it, nl, pl, uk and their cross-pairs)
• Translation quality is visible to end users: UI copy, marketing, client-facing legal documents
• Volumes are under 3-5M characters/month where the price gap isn’t yet painful
• You need CAT tool integration or a sophisticated glossary system

The Hybrid Approach - What Serious Products Do¶

Most mature multilingual products don’t commit to a single provider. The typical architecture: DeepL for European language pairs where quality matters most - Google or Azure for everything else. Plus Translation Memory (translation caching) to avoid paying twice for identical content.

This gives the best quality-to-cost ratio but requires more infrastructure: request routing by language pair, response normalization, separate billing per provider.

If API integration isn’t your use case and you just need to translate documents without writing code, services like ChatsControl have already built the API integration layer and provide a web interface: upload a document, pick the target language, get the translation.

How to Estimate Your Budget Before Launch¶

Pricing pages tell you the cost per million characters but not how many characters you’ll actually use. Here’s how to estimate correctly.

Step 1: Count characters in a representative sample

Take a typical document, page, or string and count characters. Python: len(text). Count what actually goes into the API call - including markup if you’re not stripping it.

Step 2: Multiply by volume

If your product has 10,000 active users and each translates an average of 500 characters per session, that’s 5M characters per month. If you have 100 users uploading 20-page PDFs, that could be 100K characters per document × 100 documents = 10M characters.

Step 3: Add overhead

• Retry requests on rate limit errors: +5-10%
• Test/staging environment: +10-20% of prod volume
• Separate language detection calls if you make them: small additional amount
• HTML content: add 30-50% overhead for tags (if not stripping)

Step 4: Start on the free tier and measure

All three providers offer free tiers. Run in a test environment, add logging, measure real character usage over one week - then extrapolate to a month. That number will be more accurate than any upfront estimate.

Useful endpoint: DeepL API responds to /v2/usage and returns character_count and character_limit - handy for monitoring current usage and alerting before you hit your limit.

Sources¶

1. Google Cloud Translation Pricing
2. DeepL API plans - DeepL Help Center
3. Azure Translator Pricing - Microsoft Azure
4. SimpleLocalize: How much does AI translation cost? DeepL, Google, OpenAI compared (2026)
5. Intlpull: Machine Translation Accuracy 2026 - DeepL vs Google vs ChatGPT
6. BuildMVPFast: Translation API Pricing Comparison (June 2026)
7. Taia: DeepL vs. Google Translate vs. Microsoft Translator (2026)
8. Langbly: Cheapest Translation API 2026