In Microsoft Azure, when using Azure OpenAI Service, the available (allowed) models are based on what Microsoft has approved for enterprise and responsible AI use. Here’s a general overview of the allowed models as of early 2024:

Allowed Model Families in Azure OpenAI Service

1. GPT-4

gpt-4

gpt-4-32k

gpt-4-turbo (more efficient, cheaper version of GPT-4)

2. GPT-3.5

gpt-35-turbo (also written as gpt-3.5-turbo)

3. Codex

code-davinci-002 (for code generation tasks)

code-cushman-001 (lighter Codex model)

4. Embedding Models

text-embedding-ada-002 (used for vector embedding, semantic search, etc.)

5. DALL·E Models (image generation)

dall-e (for image generation and editing)

Model Deployment in Azure

In Azure, you deploy a model to a resource (like a specific region), and then use an endpoint to make API calls.

Not all models are available in every region.

You can choose the model version when creating a deployment (e.g., gpt-35-turbo vs. gpt-4).

Model Naming Convention (Azure)

Azure uses slightly different model names than OpenAI. Examples:

OpenAI’s gpt-4 = Azure’s gpt-4

OpenAI’s text-embedding-ada-002 = Azure’s text-embedding-ada-002