GPT-4 Comes to Azure OpenAI Service

The latest generation Large Language Model (LLM), GPT-4, is now available in the Azure OpenAI Service. The GPT-4 model was trained against a much larger data set than the GPT-3 models making the model more natural in responses with more accurate responses. The GPT-4 model in the Azure OpenAI Service allow the existing model to then integrate with customer-specific data that never leaves the customer tenant. The combination brings the power of GPT-4 against business data allowing chat experiences, rapid review of materials, code analysis, and more. While the Azure OpenAI Service is Generally Available today, the GPT-4 model is in private preview. Customers need to submit their business case to Microsoft for review and approval before the GPT-4 models can be accessed. Details on the GPT-4 capabilities in the Azure OpenAI Service can be found here. 

Why This Matters

  • For many customer scenarios, the existing GPT-3 and GPT-3.5 models are often sufficient to address the business needs. However, more complex scenarios will perform better with the GPT-4 models.
  • GPT-4 can process significantly more data than GPT-3.5 so scenarios where very large amounts of data are needed, the GPT-4 models may do better. That said, many scenarios may need things like the Azure Cognitive Search to help parse the data before it is tokenized for the GPT-4 model. 
  • The GPT-4 model is the foundation for the Bing Chat and Copilot features Microsoft has announced.