Recently, ChatGPT has made two key updates to model selection and data connections: the model selector can “hold” more, and Deep Research has begun to support connecting external knowledge bases. This article takes the shortest path to help you understand what problems these new features solve and where to enable them.
The model selector gets stronger: commonly used models are back, and you can also “show more models”
In the new version of ChatGPT, some commonly used models have returned to the model selector for paid users, so you no longer need to hunt for entry points when switching. More importantly, the ChatGPT web app has added a “Show more models” toggle. After turning it on, more optional models will appear in the selector, making it easier to choose by task: a steadier one for writing, a stronger one for reasoning, and a faster one for lightweight tasks.
Note that updates like this are usually rolled out gradually; if you can’t see a certain model in ChatGPT yet, it’s often not an operation issue—it just hasn’t reached your account or region.
How to enable “Show more models” on the ChatGPT web app
After opening the ChatGPT web app, go to Settings, find the options related to model display, and toggle “Show more models” on. Return to the chat interface and open the model selector—you’ll find the list of available options noticeably longer, and the switching path more straightforward.
It’s recommended to keep a “default choice” for each common scenario: for example, use one stable model for everyday Q&A, and temporarily switch to a model that’s better at reasoning for complex analysis. This way, you won’t interrupt your workflow in ChatGPT by constantly trying models.
Deep Research supports connecting external sources: cloud drives and code repositories become more useful
Another more practical update is that ChatGPT’s Deep Research has begun to support connecting external data sources, including Dropbox, OneDrive/SharePoint, GitHub, and more. The most obvious change is that when you do research, write reports, or organize projects, you no longer need to repeatedly download and upload piles of files—after you authorize it, ChatGPT can read materials from these sources to compile and summarize them.
This capability is also being rolled out in batches, and some regions may not have access yet. If you don’t see the relevant connection entry in Deep Research, first check whether you’re in a Team/Enterprise workspace and whether the feature has been made available in your region.
Three reminders before use: quotas, privacy, and controllable “Memory”
First, capabilities such as video/screen sharing and image uploads are also being rolled out gradually in ChatGPT mobile’s advanced voice, but there are usually daily usage limits or message quotas, making them suitable for scenarios where you “need on-the-spot, on-screen explanations.” Second, before connecting external data, try to authorize only the necessary folders and repositories to reduce exposure. Third, ChatGPT’s “Memory” feature can be managed and turned off separately in Settings; if you don’t want conversations used for personalization, remember to disable the relevant options or use Temporary Chat.
