OpenAI Launches ChatGPT API

While the davinci-003 model was an improvement from GPT-3, it lacked the fine-tuning conversational capabilities of ChatGPT that many people have been looking for.

Emphasis on Shopping

A number of big vendors have already had the ChatGPT API: Snapchat's My AI feature, Shopify, Instacart, and Quizlet are some examples. OpenAI's blog release post shows an obvious emphasis on e-commerce.

It's Now Called gpt-3.5-turbo

The ChatGPT model is now called gpt-3.5-turbo. OpenAI will be releasing updated models regularly which may affect the format of API calls but developers will have access to the latest models.

It's Much Cheaper

The cost has reduced by 90%. It is really 10 times cheaper than GPT-3. This is really important, as the rising cost of the most expensive Davinci model when creating chat applications and crunching through tokens was a concern.

It's the New Go-To Model

OpenAI is pushing gpt-3.5-turbo as the model to use for many tasks people are using Davinci for, not just chat applications: composing emails, coding in Python, providing answers to document-based queries, designing conversational AI agents, offering tutelage in various subjects, translating between languages, and bringing video game characters to life.

Interesting API Changes

There is an important change in the API. ChatGPT now consumes messages with metadata, not just unstructured text, presented in a new format called Chat Markup Language (ChatML). Messages are segregated into three types: system, user, and assistant. This helps shore up security vulnerabilities from prompt injection attacks and allows for future development.

No Fine-Tuning

As of the launch, no fine-tuning of the model is available. It remains to be seen whether this will be coming. API usage data is no longer used for model training unless you opt-in.

Whisper API

Whisper, the speech-to-text model open-sourced in September 2022, has also been released as an API. Whisper is a multi-language speech recognition model, so providing a voice front-end interface with ChatGPT responses is now a nice option.

Conclusion

The new gpt-3.5-turbo model is not only cheaper but also offers more versatility beyond just chat applications and looks to be the model of choice now for most apps. The API now consumes messages with metadata in ChatML format, offering improved security. Despite the reduced cost, token management remains important.