top of page

OpenAI GPT-4 Turbo : The difference between GPTs and Assistants

Updated: Nov 10, 2023

OpenAI GPT-4 Turbo is a newly upgraded model based on GPT-4.

  • It supports longer context lengths - up to 128K tokens. (Claude 2 by Anthropic — 100K tokens)

  • Knowledge updated as of April 2023.

  • New modes such as image, voice and video functions have been added.

  • Enable retrieval to incorporate external knowledge sources.

  • Allows customization through fine-tuning and customizing models.

  • Dramatically reduced pricing - 3x cheaper on tip and 2x cheaper on completion.

Token

And the focus of this article’s discussion—GPTs and Assistants. But what are each of them, and how do they differ?

Here is a comparison of OpenAI GPT-4 Turbo’s GPTs and Assistants:


GPTs

GPTs are individual instances of ChatGPT configured for specific purposes. Although they still retain all the training content of ChatGPT, they have been "branched" through additional parameter settings to serve specific purposes. Two examples were mentioned in the OpenAI DevDay keynote speech: CanvaGPT and ZapierGPT.

  • CanvaGPT has the chat functionality of ChatGPT, but with added features specific to Canva users. You can describe your needs in natural language, and AI can show you the graphical examples you propose. With one click, you can open it in Canva and start working from the provided base design.

  • ZapierGPT also operates through natural language, and the AI ​​chatbot can query Zapier and perform the actions you want. For example, by allowing ZapierGPT access to your calendar, it can check for conflicts on specific dates and even automatically send a Slack message to your co-worker letting you know you can't make it to their child's dance recital.

OpenAI announced that it will open a GPTs store where users can browse and share public GPTs, or choose to keep them privately for personal use.


GPTs


Assistants

Assistants are similar to GPTs and also allow you to use ChatGPT content to create customized chatbots. The difference is that Assistants are aimed at developers and have no real user interface (except for OpenAI Playground).


While developers have been able to create these chatbots before, it required more knowledge of the AI ​​field and technologies such as LangChain. Assistants allow developers to focus on building the bots they want by taking care of the tedious and difficult parts.


The Assistants API allows developers to make API calls to create assistants, generate new conversation threads and attach messages to these threads. Developers can upload files for the assistant to use to better respond to user messages. Users can also upload files within their conversation thread, which will remain within the context of that thread without affecting the general operation of the assistant itself.


Assistants can also use different tools and functions. Currently (as of this writing) the tools are limited to code interpreters and retrieval. Code interpreters have a wide range of uses, they allow AI to write code and execute it. Not only for testing code, but also for actually producing something based on a user's request, such as translating an uploaded PDF file into a new PDF in a different language, or cleaning up the spreadsheet columns and creating a new spreadsheet that only contains information about a certain country. Calculation table.


The Assistants API is very intuitive, making it easy for developers to quickly generate chatbots for specific use cases and maintain communication context with specific users and assistants through threads.


Assistants


Difference

If the above information is not clear enough, I think the difference between the two is actually very simple. While they have many similarities in their end goal (making a custom chatbot), their approaches and intended use cases differ.


GPTs have a user-friendly interface and can be easily created directly from ChatGPT itself, and can be easily viewed and shared through the GPTs Marketplace, while Assistants are fully available through APIs and developers can easily integrate them into their own code bases.



Conclusion

These new features will profoundly impact the way people interact with ChatGPT, and as these new features mature, the community developing new chatbots will further grow to give us unprecedented capabilities!

As a developer, these new announcements get me very excited and inspire me to start building things with these new tools that OpenAI has to offer.

Comments


bottom of page