# OpenAI Upgrades Custom GPTs: A Detailed Summary
If you’ve been testing out custom GPTs within ChatGPT, there’s thrilling news: OpenAI now permits users to choose from all available models when constructing or operating a custom GPT, rather than just GPT-4o. This article explores what custom GPTs are, the new features offered, and some limitations to keep in mind.
## What are Custom GPTs?
Custom GPTs are personalized variations of ChatGPT that incorporate specific directives, additional knowledge, and a range of capabilities. OpenAI launched this feature in late 2023, with the goal of improving the user experience by facilitating more customized interactions.
For example, if you regularly utilize ChatGPT for social media SEO, you can develop a custom GPT that saves your specific prompts, settings, and behaviors. This means you won’t need to re-enter your instructions each time you use the model. Users have crafted custom GPTs for a myriad of purposes, including assisting students with studying or swiftly retrieving information from transcripts. Furthermore, there is a feature similar to an app store where users can browse and use GPTs developed by others.
## What’s New?
In the past, users were limited to building custom GPTs using GPT-4o. Now, the range has broadened to include several models: GPT-4o, o3, o4-mini, o4-mini-high, GPT-4.5, GPT-4.1, and GPT-4.1-mini.
OpenAI has also rolled out a “Recommended Model” option, which offers a default starting point for GPT creators while allowing them to switch models as required. According to OpenAI, “When the user initiates a new conversation with the GPT, the recommended model will be employed. If the model is not accessible to the user, a similar model may be automatically chosen.”
## Limitations to Consider
There are two prominent limitations with this new feature. First, users can only alternate between different models if their custom GPT does not utilize Actions (such as API calls, web browsing, or code interpreter features). If your GPT depends on Actions, you will be confined to a narrower selection of supported models. OpenAI has mentioned that they are addressing this limitation in the near future.
Second, the new model selection feature is at present accessible only to Pro, Plus, and Team users. OpenAI has indicated that Enterprise and EDU users will soon have access to these capabilities.
## Conclusion
The ability to select from a broader array of models when creating custom GPTs marks a considerable update since the feature’s introduction. Whether this improvement will result in heightened adoption of custom GPTs remains to be seen.
Have you been utilizing custom GPTs? Share your experiences or ideas for inventive applications in the comments.
