Azure AI Translator 2025-05-01-preview

Azure AI Translator 2025-05-01-preview is our latest cloud-based, multilingual, neural machine translation service. The Text translation API enables robust and scalable translation capabilities suitable for diverse applications.

The Translator service is an optimal solution for managing extensive multilingual content. It easily integrates with your applications and workflows through a single REST API call and supports multiple programming languages. Translator supports over 100 languages and dialects, making it ideal for businesses, developers, and organizations seeking to seamlessly integrate multilingual communication.

What's new for 2025-05-01-preview?

• LLM choice. By default, Azure AI Translator uses neural Machine Translation (NMT) technology. With the newest preview release, you now can optionally select either the standard NMT translation or Large Language Model (LLM) models—GPT-4o-mini or GPT-4o. You can choose a large language model for translation based on factors such as quality, cost, and other considerations. However, using an LLM model requires you to have an Azure AI Foundry resource. For more information, see Configure Azure AI resources

• Adaptive custom translation. You can provide up to five reference translations or translation memory datasets to enable an LLM model to perform few-shot translations in a similar style and tailored to your needs.

• Tone variant translations. Use generative AI LLMs translate text across multiple tonal categories—formal, informal, and neutral—ensuring precise contextual adaptation.

• Gender-specific language translations. Apply generative AI LLMs for the linguistic transformation of text according to specified gender parameters—male, female, and neutral—to ensure targeted stylistic and semantic alignment.

Language support

The languages supported for LLM and Adaptive custom translations are listed in the Translation section of our Language support page.

Authentication

The 2025-05-01-preview API supports both the resource API key and Microsoft Entra ID authentication. For your information, see Authorization and authentication

NMT Base URLs

Requests to Translator are, in most cases, handled by the datacenter that is closest to where the request originated. If there's a datacenter failure when using the global endpoint, the request may be routed outside of the geography.

To force the request to be handled within a specific geography, use the desired geographical endpoint. All requests are processed among the datacenters within the geography.

✔️ Feature: Translator Text

Switzerland service endpoints

Customers with a resource located in Switzerland North or Switzerland West can ensure that their Text API requests are served within Switzerland. To ensure that requests are handled in Switzerland, create the Translator resource in the Resource region Switzerland North or Switzerland West, then use the resource's custom endpoint in your API requests.

LLM processing

When you deploy a large language model (LLM), the configuration options you choose—global, data zone, or regional—directly impact and determine the specific ___location in which your data is processed. Therefore, your selections during setup play a significant role in defining the geographical boundaries for how and where the model processes your information.

Service limits

The amount of computing resources you provide influences translation latency when you use generative AI large language models. By adjusting the capacity allocated during model deployment, you can affect latency.

Pricing

• By default, translations using general NMT (Neural Machine Translation) models are billed according to the number of characters in the source text. For more information, see Azure AI Translator pricing.

• Translations using generative AI LLMs are charged according to the number of input and output tokens processed. For more information, see Azure OpenAI pricing.

Next steps