Google Cloud Translation API
Google · Ranked #1 of 7 in Translation APIs
Mature hyperscaler MT (NMT, LLM and adaptive tiers) with broad language coverage, deep SDKs and a non-expiring monthly free tier.
Hyperscaler MT at global scale
Overview
Google Cloud Translation API is Google's managed neural machine translation service, available in two editions: Basic (v2), a minimal-config REST/key-based endpoint for straightforward text translation, and Advanced (v3), an IAM-integrated service that adds document translation (DOCX/PPTX/PDF), glossaries, batch translation, custom AutoML models, and newer LLM-based translation options. It supports 130+ languages and is positioned as the breadth-and-scale leader in the translation API market, backed by Google's NMT research and global cloud infrastructure. Typical buyers are engineering and localization teams that need wide language coverage, GCP-native integration (IAM, Vertex AI, logging), and a serverless pay-per-character model rather than a managed localization vendor.
Where it wins: language coverage (far broader than DeepL's ~36 languages, including Arabic and Hindi which DeepL omits), ecosystem integration with the rest of Google Cloud, official client libraries across eight languages, and a generous 500,000-character/month free tier that resets monthly. The Advanced edition's glossary, batch, and AutoML custom-model features give enterprises real control when the base NMT model is not domain-accurate enough. The newer Translation LLM and Adaptive Translation tiers signal Google moving to close the quality gap with LLM-based competitors.
Where it loses: independent benchmarks (Intento, IntlPull) consistently rank DeepL higher on output quality for European language pairs, IntlPull's Jan 2026 test found DeepL scoring 8–16 BLEU points above Google, e.g. EN→DE 64.5 vs 48.3. Reviewers repeatedly flag that standard NMT can read literally on idioms, slang, and emotional or technical text, requiring human proofreading. Cost is the other recurring complaint: at $20 per million characters the per-character price compounds quickly for high-volume apps, and custom AutoML models start at a steep $80/M. Net: a strong default for broad-coverage, GCP-integrated translation at scale, but not the accuracy leader for premium European-language output.
How this score is derived
The APIbenchmarks Index is a weighted sum of four dimensions, each scored on an absolute 0–100 reference scale. See the methodology for every mapping.
| Dimension | Score | Weight | Contribution |
|---|---|---|---|
| Documentation & DXExtensive official docs at docs.cloud.google.com cover both editions, REST/gRPC references, code samples in every supported language, and migration/quickstart guides, among the most thorough in the category. | 88 | 30% | 26.4 |
| ReliabilityBacked by a published 99.9% monthly-uptime SLA for Basic, Advanced, and AutoML editions, with tiered financial credits (10–50%) when the objective is missed. | 93 | 25% | 23.3 |
| Ecosystem & SDKsDeep native integration with Google Cloud (IAM, Vertex AI, Cloud Storage batch jobs, logging) plus eight official client libraries make it a natural fit for teams already on GCP. | 92 | 25% | 23.0 |
| AccessibilitySelf-serve signup, API-key auth for Basic, and a 500k-character/month free tier lower the barrier to entry, though Advanced features require service-account/IAM setup. | 84 | 20% | 16.8 |
| APIbenchmarks Index (ABI) | 89.5 | ||
Table 1. Derivation of the ABI for Google Cloud Translation API. Contribution = score × weight; the index is their sum.
At a glance
- Vendor
- Pricing model
- Per million chars
- Free tier
- 500k chars/mo (never expires)
- Official SDKs
- 10 languages
Pricing
| Free tier | $0 | First 500,000 characters/month free (credit applies collectively across Basic and Advanced editions) |
| NMT text (Basic v2 / Advanced v3) | $20 / 1M chars | Standard neural machine translation over the free tier; Basic over 1B chars/month is contact-sales |
| NMT document translation (Advanced) | $0.08 / page | DOCX/PPTX/PDF; max 4,000 chars or 500 words per page |
| Translation LLM | $10 in + $10 out / 1M chars | LLM-based text translation, priced separately on input and output characters |
| Adaptive Translation (LLM) | $25 in + $25 out / 1M chars | Learns from supplied examples to adapt to style/domain |
| Custom AutoML models | $80 / 1M chars (tiered down to $30) | $80/M under 250M, $60/M to 2.5B, $40/M to 4B, $30/M above; training $45/hr capped at $300/job |
Key features
- •Neural Machine Translation across 130+ languages (Basic v2 and Advanced v3)
- •Translation LLM and Adaptive Translation (LLM-based, example-conditioned) tiers
- •Document translation for DOCX, PPTX, and PDF preserving formatting
- •Custom domain models via AutoML Translation training
- •Glossaries for consistent terminology (no charge to create)
- •Batch translation of large datasets from Cloud Storage
- •Automatic source-language detection
- •REST and gRPC APIs with IAM-integrated auth on Advanced edition
Official SDKs
Strengths & trade-offs
- +Very broad language coverage (130+ languages), including Arabic and Hindi that DeepL does not support
- +Generous 500,000-character/month free tier that resets monthly and applies across both editions
- +Deep GCP integration (IAM, Vertex AI, Cloud Storage batch, logging) and 8 official client libraries
- +Advanced edition adds glossaries, document translation, batch jobs, and custom AutoML models for domain control
- +Published 99.9% uptime SLA with financial-credit remedies
- +Newer Translation LLM and Adaptive Translation tiers offer higher-quality, customizable LLM-based output
- –Output quality trails DeepL on European language pairs in independent benchmarks (8–16 BLEU points lower in IntlPull's test)
- –Standard NMT translates idioms, slang, and emotional/technical text literally, often needing human proofreading
- –Pay-per-character cost compounds fast at scale ($20/M); custom AutoML starts at a steep $80/M
- –Weak handling of formatting such as HTML/XML reported by reviewers
- –Advanced features require service-account/IAM setup rather than a simple API key
- –Limited control over the base translation model unless you invest in custom AutoML training
What developers say
G2 4.4/5 · 330 reviews
Reviewers praise ease of integration, broad language coverage, and natural output on major languages, but consistently cite cost at scale and inconsistent accuracy on specialized or idiomatic content.
“Users consistently praise the API for its ease of integration and high accuracy across a wide range of languages; the support for 100+ languages and ability to handle formatted documents with the Advanced API saves time in localization.”
Key figures
| Monthly uptime SLA | ≥ 99.9% (Basic, Advanced, AutoML) | Google Cloud Translation SLA ↗ |
| Downtime threshold | >5% server-side error rate (HTTP 500 / Internal Error) | Google Cloud Translation SLA ↗ |
| NMT price | $20 per 1M characters (over 500k/mo free) | Google Cloud pricing page ↗ |
| Document translation price | $0.08 per page (NMT) | Google Cloud pricing page ↗ |
| Translation quality (BLEU, EN→DE) | 48.3 (vs DeepL 64.5) in 500-sentence test | IntlPull machine-translation benchmark, Jan 2026 ↗ |
| Language coverage | 130+ languages (vs DeepL ~36) | Google Cloud Translation docs ↗ |
Compare Google Cloud Translation API head to head
Sources
- https://cloud.google.com/translate/pricing
- https://cloud.google.com/translate/sla-20191015
- https://cloud.google.com/translate/sla
- https://docs.cloud.google.com/translate/docs/reference/libraries/v3/overview-v3
- https://docs.cloud.google.com/translate/docs/languages
- https://www.g2.com/products/google-cloud-translation-api/reviews
- https://www.g2.com/products/google-cloud-translation-api/reviews?qs=pros-and-cons
- https://www.capterra.com/p/264721/Google-Cloud-Translation-AI/
- https://intlpull.com/blog/machine-translation-accuracy-2026-benchmark
Figures last verified 2026-06-27. Spotted an error? corrections@apibenchmarks.com