Microsoft has announced the release of the GPT-4o-2024-08-06 model through the Azure OpenAI Service. The AI cloud service already supports GPT-4o and is now getting the latest API. OpenAI's newest version of the model aims to streamline application development by introducing Structured Outputs, which provide consistent data formats and flexible tool integrations. The model is now accessible in all US regions and Sweden Central, supporting both Global and Regional Standard deployments.
Pricing and Cost Efficiency
The new API introduces a competitively priced structure, with input tokens costing $2.50 per million and output tokens at $10.00 per million. Microsoft says this means the model offers a 50% reduction for input tokens and a 33% reduction for output tokens compared to previous versions. Developers using the GPT-4o-2024-05-13 model are advised to switch manually to the new version to benefit from these savings and enhanced features, as no automatic upgrades are planned.
The GPT-4o-2024-08-06 API brings two forms of Structured Outputs: User-defined JSON Schema and More Accurate Tool Output (known as “Strict Mode”). User-defined JSON Schema allows developers to specify the exact JSON format for the AI to follow, while Strict Mode ensures precise function signatures for tool usage. Both features are supported by models including GPT-3.5 Turbo, GPT-4, GPT-4 Turbo, and GPT-4o models from June 2023 onwards.
Practical Applications and Benefits
With the Structured Outputs, developers can minimize the need for manual post-processing by defining a JSON Schema. The feature simplifies the creation of data payloads, enhancing the overall user experience and speeding up the adoption of AI-powered tools across various sectors. For example, a customer support chatbot can produce responses ready-made for logging and analytics, ensuring smooth data integration.
The GPT-4o-2024-08-06 model is designed for tasks requiring fast, real-time text responses, balancing cost and performance. JSON Structured Outputs guarantee 100% reliability in evaluations, ensuring outputs conform strictly to the defined schemas. The improvement bolsters the efficiency and accuracy of AI applications in numerous scenarios.
Getting Started with the New Model
To take advantage of the Structured Outputs, developers should define the JSON Schema, configure the AI model, and integrate and test the output within their applications. Both GPT-4o-2024-08-06 and GPT-4o Mini feature multimodal capabilities, with the GPT-4o Mini optimized for cost-effectiveness, making it suitable for testing and cost-sensitive applications.
Comprehensive documentation is available to help developers understand the model and its features. Azure AI Studio also offers a platform for trying out the model. While automatic upgrades to the new model are not planned, Microsoft will keep users informed of any future updates.
