Stay free if you only need up to 5,000 units per feature per month at no cost; ideal for prototyping, small projects, and evaluation.. Upgrade if you need custom-trained entity extraction, sentiment, and classification models billed for node-hours of training and prediction throughput, in addition to standard storage and infrastructure costs.. Most solo builders can start free.
Why it matters: Pricing is per-unit and can become expensive at high volumes compared to self-hosted open-source alternatives like spaCy or Hugging Face Transformers
Available from: Standard Usage
The API supports a broad range of languages depending on the feature. Entity analysis, sentiment analysis, and syntax analysis cover major languages including English, Spanish, French, German, Italian, Portuguese, Chinese (Simplified and Traditional), Japanese, Korean, and Russian. Content classification is primarily optimized for English, with expanded coverage for additional languages over time. Coverage varies by feature, so the official documentation should be consulted for the exact matrix.
Pricing is based on units, where one unit equals 1,000 characters of text. Each feature (entity analysis, sentiment, syntax, classification, entity sentiment) is billed independently per unit. Google offers a free tier of up to 5,000 units per feature per month, after which tiered pricing applies, with discounted rates as monthly volume increases.
Yes. For domain-specific entity extraction, sentiment, or classification, Google offers AutoML Natural Language (now part of Vertex AI). You upload labeled examples and Vertex AI handles model training, evaluation, and deployment. The resulting custom model is served behind a similar API and can be used alongside or instead of the pre-trained models.
The Natural Language API is a task-specific service with deterministic, structured output schemas optimized for entity extraction, sentiment, and classification. Gemini and other LLMs are general-purpose generative models that can perform similar tasks via prompting but with less predictable output structure, higher per-call cost at scale, and different latency profiles. The Natural Language API is typically preferred for high-volume, structured analytics pipelines, while LLMs are preferred for flexible, reasoning-heavy tasks.
Google Cloud provides enterprise-grade security including encryption in transit and at rest, IAM-based access control, VPC Service Controls, and compliance certifications such as SOC 2, ISO 27001, and HIPAA. However, customers must evaluate data residency, retention, and regional processing requirements against their specific compliance obligations and configure the service accordingly.
Start with the free plan — upgrade when you need more.
Get Started Free →Still not sure? Read our full verdict →
Last verified March 2026