During their first-ever DevDay, OpenAI pushed the game further. Let's dive into what developers can now accomplish.
- 2x faster
- 2.75x cheaper
- 128k context input, 4k output
- JSON mode to enforce responses in JSON format (demo here)
- Support for multiple functions calls in a single request (demo here)
- Knowledge extended through April 2023
- Ability to pass images as input (demo here)
- OpenAI API can now generate speech from text (demo here)
- Offers 6 voices (documentation)
- Supports real-time streaming (songs start playing before audio generation completion)(demo here)
- Note
⚠️ : Custom voices are not available at this time
⚠️ Available on ChatGPT GUI only ⚠️
- OpenAI now provides the ability to create your custom versions of ChatGPT, called GPTs
- GPTs can be shared with others
- OpenAI introduces the option to create an
action, enabling APIs for your GPTs (already utilized by Canva and Zapier) - OpenAI can host your OpenAPI (API!) specifications for these actions
- Actions can be used with APIs that require authentication
- OpenAI introduces Assistants (demo here)
- You can upload files to OpenAI, which will be accessible to your assistants (demo here)
- Each assistant can link up to 20 files, each with a maximum size of 512MB
- Assistant have access to various tools, including functions and code interpreter
- OpenAI introduces Threads, a way to organize conversations (demo here)
- Threads inherit the files and other configurations linked to the assistant (demo here)
- You can attach up to 10 files to a thread message