OpenAI unveiled a slew of new capabilities, reduced pricing, and increased rate limits across its developer platform at its DevDay event today. The announcements aim to significantly upgrade OpenAI’s offering for developers building AI-powered apps and services.
The headliner is GPT-4 Turbo, a more advanced version of GPT-4 that OpenAI says is more capable, cheaper, and supports a 128K context window compared to the original GPT-4’s 32K context. This enables the model to fit the equivalent of over 300 pages of text in a single prompt for more coherent conversations. OpenAI optimized GPT-4 Turbo’s performance so it can offer the model at a 3x cheaper price per input token and 2x cheaper per output token versus GPT-4.
OpenAI also introduced an Assistants API that makes building AI assistants easier by handling thread state management and letting developers call tools like code interpreter, retrieval, and function calling. This provides persistent conversations and extra capabilities without developers having to handle the complexity themselves.
On the multimodal front, OpenAI announced GPT-4 Turbo will soon support vision, enabling analyzing images in text prompts. DALL-E 3 image generation will also become available in the Images API. And a new text-to-speech API with high-quality voices was launched.
Other notable updates include reproducible outputs, log probabilities, improved instruction following, releasing Whisper v3 for automatic speech recognition, and open sourcing the Consistency Decoder to improve Stable Diffusion image generation.
For developers, all these new capabilities come with reduced pricing across the board. OpenAI cut prices by up to 4x for various models’ input and output costs. Usage limits are also doubled for GPT-4 customers. OpenAI is rolling out the new features starting today and in the coming weeks.
The OpenAI announcements aim to supercharge what developers can build with AI and make it more accessible. GPT-4 Turbo and the new modalities create opportunities for more advanced applications, while the updated pricing and limits remove barriers to usage. It’s an exciting upgrade for the developer community looking to leverage AI. With these enhanced tools, we may soon see another wave of innovative AI-powered apps, products and services.