OpenAI held its inaugural DevDay event today, unveiling a slew of new capabilities, upgrades and pricing changes for its developer platform. The announcements paint the picture of an AI company rapidly evolving its offerings to be more powerful, accessible and multifunctional. From major leaps in its foundation models to new ways to customize, OpenAI is providing developers more tools to build the next generation of AI applications.
- GPT-4 Turbo Launched, Offering 128K Context at Lower Price
OpenAI released a more advanced version of GPT-4 called GPT-4 Turbo. It handles 128K context, allowing prompts equivalent to 300 pages of text. Pricing is now 3x cheaper for input and 2x cheaper for output versus GPT-4.
- New Assistants API for Building Agent-Like Apps
The Assistants API helps developers build apps with persistent state and tools like code execution, knowledge retrieval and function calling. This unlocks more advanced AI capabilities.
- Multimodal Abilities Added, Including Vision and Text-to-Speech
GPT-4 Turbo can now process images, enabling vison-related tasks. DALL-E 3 integration allows generating images via API. New text-to-speech voices were added with two quality options.
- Improved Accuracy of Function Calling
Models are now better at calling functions correctly and can call multiple functions in one prompt. This reduces the need for multiple API calls.
- Reproducible Outputs Introduced
A new seed parameter introduces reproducible outputs from models, aiding testing and debugging. Log probabilities for tokens also added.
- GPT-3.5 Turbo Updated, Price Reductions Announced
An updated GPT-3.5 Turbo with 16K context was released. Various price decreases were announced, with GPT-3.5 Turbo input cost dropping 3x.
- New Programs for Customization and Copyright Protection
Organizations can apply to request custom model training services from OpenAI directly. Copyright Shield introduced to protect customers from infringement claims.
OpenAI's barrage of announcements at DevDay 2023 highlights the impressive pace at which artificial intelligence is progressing and solidifies their position at the forefront of AI technology. In just the span of a year, capabilities like handling longer contexts and multimodal inputs have gone from research projects to developer-ready APIs and user-facing products.
As the company continues democratizing access through more affordable pricing and customizable offerings, it's clear we're only beginning to grasp the transforming potential of AI. For developers and businesses, the possibilities to build helpful, intelligent products have expanded tremendously. As these technologies become more accessible and refined, we can expect a surge in AI-powered solutions across industries, reshaping the landscape of what's possible with artificial intelligence.