This update makes it easier than ever to build and scale with cutting-edge AI tools.
We’re excited to announce the launch of OpenAI’s latest model on Azure. The new model, officially named GPT-4o-2024-08-06, introduces innovative features aimed at enhancing developer experiences on Azure. It focuses on boosting productivity by providing Structured Outputs, such as JSON Schemas, for the new GPT-4o and GPT-4o mini models.
Emphasizing Structured Outputs
GPT-4o was initially introduced in May 2024 as OpenAI’s new multimodal model, with GPT-4o mini following in July 2024. The latest version is specifically designed to simplify generating well-defined, structured outputs from AI models. This feature is especially beneficial for developers who need to validate and format AI outputs into structures like JSON Schemas, a task that often presents challenges.
Structured Outputs solve this challenge by allowing developers to specify the desired output format directly from the AI model. With this feature, developers can define a JSON Schema for text outputs, making it easier to generate data payloads that seamlessly integrate with other systems or improve user experiences.
Use Cases for JSON
JSON Schema is crucial for defining the structure and constraints of JSON documents, ensuring they adhere to specific formats with required properties and value types. It improves data clarity through semantic annotation and acts as a domain-specific language for optimizing application needs. Development teams use JSON Schema to maintain platform consistency, enforce model-driven UI constraints, and automatically generate user interfaces. It supports data serialization, security testing, and partial validation in technical scenarios. JSON Schema also facilitates automated testing, schema inference, and machine-readable web profiles, enhancing data interoperability. It standardizes validation processes and reporting, manages external validation, and ensures data consistency within and across documents. Additionally, it can assist in customer support and timely communication.
Two Versions of Structured Outputs
Structured Outputs come in two options:
- User-defined JSON Schema: This option allows developers to specify the exact JSON Schema for the AI to follow. It is supported by both GPT-4o-2024-08-06 and GPT-4o-mini-2024-07-18.
- More Accurate Tool Output (“Strict Mode”): This version enables developers to define specific function signatures for tool use. It is supported by all models that allow function calling, including GPT-3.5 Turbo, GPT-4, GPT-4 Turbo, and GPT-4o models from June 2023 onwards.
Technical Guidance for Using Structured Outputs
To help you get started with Structured Outputs, we suggest the following approach.
Getting Started with Structured Outputs
- Define Your JSON Schema: Identify the structure you want your AI outputs to follow, including required fields, data types, and other constraints.
- Configure the AI Model: Utilize the Structured Outputs feature to specify your JSON Schema within the API call, ensuring that the AI output aligns with your defined structure.
- Integration and Testing: Integrate the output into your application or system and thoroughly test it to ensure it complies with your JSON Schema.
Example Use Case: Automating Customer Support
Consider developing a customer support chatbot that must generate responses in a specific format for logging and analytics. With Structured Outputs, you can define a JSON Schema that includes fields like responseText, intent, confidenceScore, and timestamp. This guarantees that every response from the chatbot is correctly formatted, simplifying the processes of logging, analyzing, and taking action.
Example API Call
Here’s a sample API call to demonstrate how to use Structured Outputs:
Pricing
Pricing details for this feature will be available soon. Please bookmark the Azure OpenAI Service pricing page for updates.
Explore the Future of AI
We’ve recently introduced several new models, and we know it can be overwhelming to stay updated. This surge in activity is aimed at fostering developer innovation. Each new model offers unique capabilities and enhancements, enabling you to create even more powerful and versatile applications.
The introduction of this new model feature for GPT-4o and GPT-4o mini represents a major milestone in our continuous efforts to advance AI capabilities. We’re eager to see how developers will utilize these features to build innovative and impactful applications.
Stay tuned for more updates and prepare to explore the future of AI with these new developer features for GPT-4o and GPT-4o mini. Begin experimenting in the Azure OpenAI Playground.