OpenAI Unveils GPT-4 Turbo with 128K Context and Enhanced Developer Tools at DevDay

On November 6, 2023, OpenAI announced a series of significant updates and new features at their annual DevDay. The headline news includes the release of GPT-4 Turbo with an extended 128K context window, lower pricing, and enhanced multimodal capabilities. Additionally, they introduced the Assistants API, DALL·E 3 API, and improvements to function calling in the platform.

GPT-4 Turbo: More Capable, Cheaper, and with a 128K Context Window

The new GPT-4 Turbo model is a major step forward from its predecessor. Here are the key changes:

• Extended Context Window: GPT-4 Turbo now supports a context window of up to 128K tokens, which translates to more than 300 pages of text in a single prompt. This is a significant improvement for applications that require handling large amounts of information at once.
• Cost Reduction: The pricing for input tokens has been reduced by 3x, and output tokens by 2x compared to the standard GPT-4 model. This makes it more accessible for developers working on budget-constrained projects.
• Improved Knowledge Base: GPT-4 Turbo has knowledge of world events up to April 2023, making it more current and relevant for a wide range of applications.

GPT-4 Turbo is currently available in preview mode for all paying developers by using the gpt-4-1106-preview endpoint. OpenAI plans to release a stable production-ready version in the coming weeks.

Assistants API: Building Smarter AI Apps

The new Assistants API is designed to simplify the process of building assistive AI applications. Key features include:

• Goal-Oriented Assistants: Developers can create assistants with specific goals and behaviors, making it easier to tailor the AI to specific use cases.
• Model and Tool Integration: The API allows for seamless integration with other models and tools, enabling more complex and versatile applications.

This update aims to reduce the complexity of building sophisticated AI apps, allowing developers to focus on higher-level design and functionality.

Multimodal Capabilities: Vision, Image Creation, and Text-to-Speech

OpenAI is expanding its platform's multimodal capabilities with:

• Vision API: This new feature allows GPT-4 Turbo to process and understand images alongside text. This opens up possibilities for applications that require both visual and textual analysis.
• DALL·E 3 API: The latest version of DALL·E, now available as an API, enables developers to generate high-quality images based on text prompts. This is a significant upgrade from previous versions, offering more detailed and contextually relevant images.
• Text-to-Speech (TTS): While not explicitly detailed in the announcement, TTS capabilities are mentioned as part of the broader multimodal enhancements.

Function Calling Updates

Function calling has been improved to enhance the efficiency and accuracy of AI interactions:

• Parallel Function Calls: Users can now request multiple actions in a single message. For example, "open the car window and turn off the A/C" can be processed in one go, reducing the need for multiple roundtrips with the model.
• Improved Accuracy: GPT-4 Turbo is more likely to return the correct function parameters, making it easier to integrate with external APIs and applications.

Rollout and Availability

OpenAI will begin rolling out these new features to customers starting at 1pm PT on November 6. Developers can start exploring the new capabilities immediately by using the provided endpoints and API documentation.