Microsoft Unveils New Azure AI Translator API with Generative AI Capabilities in Public Preview

Microsoft Launches Next-Generation Azure AI Translator API with Generative AI Features

Microsoft announced the public preview of its new Azure AI Translator API, a major leap forward in enterprise translation services. This API empowers businesses, developers, and organizations to select between general neural machine translation (NMT) and powerful generative AI large language models (LLMs) on a per-request basis, providing unprecedented flexibility, control, and customization for global language needs.

Advanced Translation Features for Modern Workflows

The revamped Azure AI Translator API enables users to:

• Switch Between Translation Engines: Choose between traditional NMT for fast, deterministic results or generative AI (including GPT-4o and GPT-4o-mini during preview) for nuanced, natural, and context-adaptive translations.

• Customize Tone and Gender: The API can distinguish and translate text in different tones (formal, informal, neutral) and genders (male, female, neutral), addressing a broad range of localization and inclusivity requirements.

• Support Diverse Workflows: Translate with human review for localization or rapid automated translation for real-time scenarios—all within a single workflow. Users can even combine translation models in one API call.

• Few-Shot References: Provide up to five reference translations for “few-shot” learning, letting the API mimic preferred linguistic style or terminology.

How the New API Works

For businesses, the new service model means deciding—request by request—based on their specific needs:

• General NMT: Use for real-time, low-latency translation.

• Generative AI LLMs: Best for quality, nuanced translation in creative, document-heavy, or reviewed scenarios (requires deploying LLM instances in Azure AI Foundry).

• Flexible API Usage: Parameters are now supplied in the request body, released from the old method of passing them via query string, making the API more extensible and secure.

Technical Details & Pricing

• Supported Models: General NMT, GPT-4o, and GPT-4o-mini (preview)

• Authentication: Works with both Azure resource key and Entra ID

• Pricing: NMT translations are charged per-character, while generative AI translations are billed per-token (see official pricing links)

• Service Limits: NMT supports up to 1,000 elements per request; LLMs support up to 50 elements and 5,000 characters each.

Sample Use Cases

• Localize apps and websites: Instantly adapt communications with tone and style guidelines.

• Document translation: Automate large-scale content localization for global business.

• Inclusive language: Ensure communication consistency across gendered/neutral languages and markets.

Future of AI-Driven Language Technology

The API’s open preview enables customers to experiment and innovate with multi-model approaches, laying the groundwork for future enhancements—including AI-driven document image translation and batch multi-format support rolling out in 2025.

Microsoft’s new Azure AI Translator API signals an industry-wide move toward more intelligent, adaptable, and inclusive machine translation, keeping businesses agile in a globalized, multilingual world.