If you're evaluating the Google Translate API for a project, the pricing page on Google Cloud can be confusing. There are multiple API versions, separate pricing for different translation models, character counting rules that aren't obvious, and a free tier that works differently than you'd expect.
I spent a few hours digging through the actual pricing documentation, testing the billing behavior, and comparing it against what you'd pay with other providers. This guide covers everything you need to make a real cost estimate.
The short version
Google Translate API charges $20 per million characters for both the Basic (v2) and Advanced (v3) editions. You get 500,000 characters per month free. After that, every character costs $0.00002.
That's the simple answer. The detailed answer has more layers.
Google Translate API pricing by version
Google offers two main API versions, plus add-ons for custom models and newer AI-based translation. Here's what each costs:
Cloud Translation Basic (v2)
This is the original Google Translate API that most developers know. It translates text between 130+ languages using Google's Neural Machine Translation models.
| Volume | Price per 1M characters | Per-character cost |
|---|---|---|
| First 500K/month | Free | $0 |
| 500K to 1 billion/month | $20 | $0.00002 |
| Over 1 billion/month | Contact sales | Custom |
The v2 API is straightforward. You send text, you get translations back. No glossary support, no batch mode, no document translation. Just text in, text out.
Cloud Translation Advanced (v3)
The v3 API adds glossaries, batch translation, document translation, and access to custom-trained models. For standard NMT text translation, the price is identical to v2:
| Feature | Price |
|---|---|
| NMT text translation | $20 per 1M characters |
| Document translation (NMT) | $0.08 per page |
| Document translation (custom model) | $0.25 per page |
| Free tier | 500K characters/month (shared with v2) |
If you only need text translation, there's no pricing advantage to v3. The difference is in features, not cost.
AutoML Translation (custom models)
If you train your own translation model through Google's AutoML, the pricing jumps significantly:
| Volume | Price per 1M characters |
|---|---|
| First 500K/month | Free (separate $40 credit) |
| 500K to 250M | $80 |
| 250M to 2.5B | $60 |
| 2.5B to 4B | $40 |
| Over 4B | $30 |
Training a custom model costs $45 per hour, capped at $300 per training job. Most companies won't need this unless they have very specific terminology requirements that glossaries can't handle.
Adaptive Translation (newer, AI-based)
Google recently added AI-powered translation options through the v3 API. These charge for both input and output characters:
| Model | Input cost per 1M chars | Output cost per 1M chars | Effective total |
|---|---|---|---|
| Adaptive Translation | $25 | $25 | ~$50 per 1M chars |
| TextTranslation LLM | $10 | $10 | ~$20 per 1M chars |
The Adaptive Translation model is expensive. At $50 effective per million characters, it's 2.5x the cost of standard NMT. The TextTranslation LLM model lands at roughly the same cost as NMT, but charges both ways. Neither model is covered by the free tier.
The free tier: what you actually get
Google's free tier is 500,000 characters per month. A few things to know about it:
- It's permanent. Unlike the $300 GCP trial credit (which expires after 90 days), the 500K monthly allowance stays.
- It covers both v2 and v3 NMT translation. The credits are shared, not separate.
- It does not cover document translation, Adaptive Translation, or the TextTranslation LLM model.
- Unused credits don't roll over to the next month.
- AutoML gets a separate free tier: 500K characters per month with a $40 credit equivalent.
For context, 500K characters is roughly 80,000-100,000 words, or about 200-250 pages of text. That's enough for testing and small projects, but any production application will exceed it quickly.
Character counting: where costs sneak up
This is where most developers underestimate their Google Translate bill. The character counting rules have some gotchas:
HTML tags count as characters. If you send HTML-formatted text using format=html, every character in every tag gets billed. A paragraph like <p class="content">Hello</p> costs you 26 characters, not 5. Use format=text when you don't need HTML processing to avoid paying for markup.
Whitespace counts. Spaces, tabs, newlines. They all count toward your character total.
Empty requests cost 1 character. If you accidentally send an empty string to the API, Google bills it as 1 character. This matters if you have a bug that sends empty requests in a loop.
Multi-language requests multiply. If you translate 5,000 characters into 3 languages in a single batch request, you're billed for 15,000 characters (5,000 x 3).
Characters mean Unicode code points. Most Latin text counts as expected (1 character = 1 code point). But some emoji and CJK characters can be 1-2 code points, and combined emoji sequences can be more.
What does Google Translate actually cost per month?
Let's put real numbers to common usage scenarios. All calculations assume standard NMT translation (v2 or v3) after the free tier is exhausted:
| Monthly volume | Google Translate cost | Annual cost |
|---|---|---|
| 500K characters | Free | $0 |
| 1M characters | $10 | $120 |
| 5M characters | $90 | $1,080 |
| 10M characters | $190 | $2,280 |
| 25M characters | $490 | $5,880 |
| 50M characters | $990 | $11,880 |
| 100M characters | $1,990 | $23,880 |
(The free 500K is subtracted from each tier, so 5M characters costs $20 x 4.5 = $90, not $100.)
For a mid-size e-commerce platform translating product catalogs into 5 languages, 25M characters per month is realistic. That's $490/month or nearly $6,000 per year on translation alone.
Google Translate API vs. alternatives: cost comparison
Google isn't the only option, and it's not the cheapest. Here's how the major translation APIs compare at different volumes:
| Monthly volume | Google Translate | DeepL API Pro | Amazon Translate | Langbly |
|---|---|---|---|---|
| 1M characters | $10 | $25.49 | $15 | Free (within 500K tier) |
| 5M characters | $90 | $105.49 | $75 | $19 |
| 25M characters | $490 | $505.49 | $375 | $69 |
| 100M characters | $1,990 | $2,005.49 | $1,500 | $199 |
At 5 million characters per month, you'd pay $90 with Google or $19 with Langbly. That's a 79% difference. At 100 million characters, the gap widens to 90%.
DeepL charges the same $20/1M rate as Google, plus a $5.49 monthly base fee. Amazon Translate is slightly cheaper at $15/1M, but has a 1,000-character minimum per request that can inflate costs for short strings. For a detailed provider comparison, see our translation API pricing comparison.
Why is there such a price gap?
Google, DeepL, and Amazon all run custom-trained Neural Machine Translation models on dedicated GPU infrastructure. They've invested billions in training data and compute. The per-character pricing reflects years of R&D amortization and enterprise-grade infrastructure costs.
Newer translation APIs like Langbly use a different architecture. Instead of training and serving custom models, they leverage advanced AI models that handle translation as one of many capabilities. The shared infrastructure and competitive AI inference market drives per-character costs down to roughly $0.10 per million characters, which is why the end-user pricing can be so much lower.
The trade-off is latency. Google Translate responds in 50-200ms. Context-aware translation approaches take 500ms-3 seconds depending on text length. For batch processing, localization pipelines, and backend API calls, this rarely matters. For real-time chat translation, it might. Our cost reduction guide covers this trade-off in more detail.
How to reduce your Google Translate costs
If you're staying on Google Translate, a few practical steps can reduce your bill:
Use format=text instead of format=html. Strip HTML before sending text to the API. This alone can cut your billed characters by 20-40% depending on how tag-heavy your content is.
Cache translations. If you're translating the same strings repeatedly (UI labels, product categories, navigation menus), cache the results. Google doesn't offer built-in caching, so you'll need to build this yourself with Redis or a database lookup.
Batch smartly. The API accepts up to 2,048 strings per request. Batching reduces HTTP overhead but doesn't change the per-character cost. Where batching helps is in reducing the chance of empty or near-empty requests that waste API calls.
Detect language separately. Language detection is free when bundled with a translation call. If you're using detect-only calls, make sure you're not accidentally paying for unnecessary translation.
Monitor billing actively. Set up Google Cloud budget alerts. Translation costs can spike unexpectedly if a feature change increases the volume of text you're sending, or if a bug sends duplicate requests.
Switching from Google Translate: what to consider
If Google's pricing doesn't work for your volume, switching is easier than you'd think. Langbly's API is 100% compatible with the Google Translate v2 API. Same endpoint structure, same request/response format, same error codes. The migration is literally changing one URL:
// Before
POST https://translation.googleapis.com/language/translate/v2
// After
POST https://api.langbly.com/language/translate/v2
Your existing code, SDKs, and integrations keep working. No schema changes, no new authentication flows. We also have official Python and Node.js SDKs if you want a dedicated client. For a step-by-step walkthrough, check the API documentation.
Langbly includes a few things Google doesn't offer at the API level: built-in 7-day response caching, no credit card required for the free tier, and context-aware translation that handles idioms and register better than NMT. The full comparison covers the differences.
Frequently asked questions
Is the Google Translate API free?
Partially. You get 500,000 characters per month free, which resets monthly and never expires. Beyond that, it's $20 per million characters. New Google Cloud accounts also get a $300 trial credit that covers all GCP services, but it expires after 90 days.
How much does Google Translate API cost per word?
At $20 per million characters, the per-word cost depends on your average word length. For English text (average ~5 characters per word plus a space), the cost works out to roughly $0.12 per 1,000 words, or $120 per million words.
Is Google Translate API v2 or v3 cheaper?
They cost the same for standard NMT text translation: $20 per million characters. The v3 API offers additional features (glossaries, batch mode, document translation) but no pricing advantage for basic text translation.
Does Google offer volume discounts for translation?
Only for volumes above 1 billion characters per month, and only through sales negotiations. There are no published volume tiers or committed-use discounts for Cloud Translation.
What's cheaper than Google Translate API?
Several alternatives cost less per character. Langbly ranges from $1.99 to $3.80 per million characters depending on your plan. Amazon Translate is $15 per million. Microsoft Translator is $10 per million. See our translation API comparison for the full breakdown.
Do spaces and HTML tags count as characters?
Yes. Spaces, whitespace, and HTML tags all count toward your billed character total. Use format=text and strip HTML before sending to reduce costs.
Bottom line
Google Translate API pricing is simple on the surface ($20/1M characters) but the real cost depends on how you handle HTML, how many target languages you translate into, and whether you're caching effectively. For volumes under 500K characters per month, it's free. For anything above that, it gets expensive fast compared to newer alternatives.
If your monthly bill is above $50, it's worth running a quick test with a cheaper provider. A drop-in compatible API like Langbly lets you compare quality and cost without rewriting any code.