OpenAI has announced several updates to its Platform for developers. One major update is the ability to test voice outputs. Developers can now use a microphone to dictate inputs and also listen to the audio output while reading the transcribed version generated by the audio model. This feature is particularly useful for users who do not have access to ChatGPT Plus.
Another update is the introduction of the Evaluations feature, which is accessible in the Dashboard tab. This tool allows developers to input multiple user scenarios to validate how a model’s prompt performs against a benchmark.
Interestingly, a similar feature has been available on Anthropic Workbench for some time. Anthropic’s version also enables developers to create specific test cases for their prompts.
It remains to be seen whether OpenAI’s Evaluations feature will become available through the API. If so, developers could use it as part of a continuous integration process. However, there has been no official announcement or clarification regarding this possibility.