Default Version of GPT-4o Updated to Latest GPT-4o Model

In a significant development for the AI community, OpenAI has announced that the default version of the GPT-4o model will be updated to the latest version, gpt-4o-2024-08-06, starting October 2nd, 2024. This update marks a major enhancement to the capabilities and efficiency of the GPT-4o model.

Key Improvements

The new version of GPT-4o introduces several key improvements:

• Cost Efficiency: The update brings a 50% reduction in the cost of input tokens and a 33% reduction in the cost of output tokens, making it more economical for users to process data and generate responses.
• Structured Outputs: The gpt-4o-2024-08-06 model supports structured outputs, enabling more complex and organized responses from the model. This feature enhances the model's ability to handle multi-step tasks and provide more coherent and structured responses.
• Increased Output Tokens: The maximum output tokens have been increased from 4,096 to 16,384, allowing for more extensive and detailed responses.

Multimodal Capabilities

GPT-4o, with its "Omni" designation, is a multimodal model that can handle text, images, and audio inputs and outputs. This integration allows for more natural and intuitive interactions with users, making it suitable for a wide range of applications including real-time conversations, Q&A, text generation, and more.

Performance and Capabilities

The GPT-4o model boasts several advanced features:

• Context Window: It supports a context window of up to 128,000 tokens, enabling it to maintain coherence over longer conversations or documents.
• Vision and Audio: The model has advanced vision and audio capabilities, outperforming its predecessors in these areas. It can respond with AI-generated voices that sound human and has a rapid audio input response time of approximately 320 milliseconds.
• Non-English Languages: GPT-4o shows superior performance in non-English languages compared to other models, making it a robust choice for global applications.

Deployment and Availability

The update will be seamless for users currently using the gpt-4o model parameter in their API requests. However, users who prefer to stick with the older version (gpt-4o-2024-05-13) can do so by explicitly specifying it in their API requests.

The gpt-4o-2024-08-06 model is available for standard and global-standard deployments in all US regions and Sweden Central. Additionally, the gpt-4o mini model, which is a smaller and more affordable version, is available for deployment in various regions, including East US, Sweden Central, and West US.

Fine-Tuning and API Updates

Alongside the model update, OpenAI has also introduced GPT-4o fine-tuning in public preview for Azure OpenAI in North Central US and Sweden Central. This feature allows for more customized model performance tailored to specific use cases.

The latest API release, 2024-07-01-preview, includes batch API support and other enhancements, further expanding the capabilities of the GPT-4o model.

This update is expected to significantly enhance the performance, efficiency, and cost-effectiveness of applications built using the GPT-4o model, making it an exciting development for developers, researchers, and businesses leveraging AI technologies.