How to use Azure openAI in Trados Studio 2024

Question

Dear all, 
 does anybody know which Azure openAi option we can use to only do some minor tests regarding the integration of LLM/Azure in Trados 2024 and machine translation? 
 There are numerous options on the Azure price list ( Azure OpenAI Service &ndash; Preise | Microsoft Azure , country Germany) and we don't know which is/are suitable. 
 Furthermore, is there anyboday who has already set up an Azure account and can quickly guide us through this process? 
 
 Thank you for your help! 
 Sabrina

Patrick Andrew Hartnett · Answer

Hi Bettina Weinmann , I'll try help you as best I can, although I haven't setup this up in Azure, there is a lot of documentation around this from MS to help create minimal walk-through underneath. 
 
 Recommended models (balance of quality/speed/cost) 
 
 Default for testing and most production: gpt-4o-mini (OpenAI or Azure OpenAI). Lowest cost, very fast, strong translation quality. 
 When quality really matters: gpt-4o. Better handling of nuance/formatting; higher cost and slightly slower. 
 Not recommended for MT: o1/o3 (reasoning-focused, slow/expensive) and 3.5 family (older, weaker quality/formatting).

Useful Links 
 
 RWS AppStore (plugin + docs/screenshots): https://appstore.rws.com/ (search &ldquo;OpenAI Provider&rdquo;) 
 Azure OpenAI access: https://aka.ms/oai/access 
 Azure OpenAI model availability by region: https://learn.microsoft.com/azure/ai-services/openai/concepts/models 
 Azure OpenAI pricing: https://azure.microsoft.com/pricing/details/cognitive-services/openai-service/ 
 Data/privacy (Azure OpenAI): https://learn.microsoft.com/azure/ai-services/openai/overview#data-privacy

Quick Azure OpenAI setup (minimal-cost) 
 
 Access and subscription
 
 Ensure your Azure subscription has Azure OpenAI enabled ( https://aka.ms/oai/access ).

Create the resource
 
 Azure Portal > Create resource > Azure AI Services > Azure OpenAI. 
 Choose an EU region that has your target model (often West Europe or North Europe; Germany regions may not have all models).

Deploy a model
 
 Go to Azure AI Foundry ( https://ai.azure.com ) or the resource in the Portal. 
 Create a Deployment for gpt-4o-mini (or gpt-4o). Give it a short deployment name, e.g., gpt4o-mini.

Get connection details
 
 From the Azure OpenAI resource: copy Endpoint (looks like https://YOURRESOURCE.openai.azure.com ), and one of the Keys. 
 Note the API version your deployment supports (for chat/completions; a recent 2024 date is fine).

Configure the OpenAI Provider in Studio 2024 
 
 Install the OpenAI Provider from the RWS AppStore. 
 Studio > Project Settings > Language Pairs > Translation Memory and Automated Translation > Add > OpenAI Provider. 
 Choose a backend:
 
 OpenAI API: paste your OpenAI API key; Base URL https://api.openai.com/v1 ; Model gpt-4o-mini or gpt-4o. 
 Azure OpenAI: select Azure; paste Endpoint (root URL only), API key, Deployment name (the one you created), and API version.

Suggested settings for translation:
 
 Temperature: 0.0&ndash;0.2 for consistent, non-creative output. 
 Ensure &ldquo;chat/completions&rdquo; is used.

Put the provider near the top of your MT list so it&rsquo;s queried during interactive translation and pre-translate.

RWS AppStore > 1. General

How to use Azure openAI in Trados Studio 2024

Top Replies