Has Model Selection Been Removed from ChatGPT Plus?
I am a ChatGPT Plus subscriber, and I can no longer find any way to select a named model for an ordinary Chat conversation.
I have checked both the ChatGPT web interface and the Windows desktop application.
On the web, there is no visible named-model selector. The controls in the message composer allow me to choose options such as Instant, Medium, or High, but those change the level of reasoning. They do not allow me to choose between GPT-5.6 Sol and GPT-6.
Could OpenAI please answer the following:
Can Plus subscribers still choose the named model used for an ordinary Chat conversation?
If so, where is that control in the current web and Windows interfaces?
Has named-model selection been replaced by reasoning-level choices such as Instant, Medium, and High?
If the underlying model is assigned automatically, where can the user verify which model is responding?
Can OpenAI change the underlying model during an existing conversation, and if so, is that change disclosed?
What does “explicit GPT-5.6 selection” mean under the current interface when no named-model selector is visible?
This is more than a labeling problem. Different models can produce noticeably different results in reasoning, tone, continuity, and instruction-following. A user cannot make an informed choice, or even determine whether a stored preference is being honored, when the interface exposes only reasoning levels while the underlying model remains unidentified.
If Plus subscribers no longer have named-model choice in Chat, please state that plainly. If we do retain that choice, please provide the current, verified steps for exercising it.
There is a non-obvious interaction, which is to click on the word “Medium”, the current reasoning effort. Click directly above the slider, where the reasoning effort has a > symbol.
Check if it just the poor user interface, and that you’ve got everything expected once discovering the right control. This is seen in web, which is kept updated.
Then check in settings your subscription plan, that it hasn’t lapsed and you still have “Plus” through to a future date.
I give a link below to OpenAI pricing and plans.
In the grid of plans is also a list of models. This shows model names per plan, without making distinction between deprecated “chat” and future “work” (alternately Codex mode, that may not have a special name after the transition).
5.6 models are still listed.
I can’t tell specifically if “Plus” has lost 5.6 models on my own “Pro” subscription, because a Pro plan says “access to legacy models” (even though it also stops at 5.6). The feature documentation is wrong when it offers that Pro has any access to legacy models older than 5.6, and the listing even has gpt-5-thinking mini that is nowhere to be seen.
If nothing else can give you access, and other plus users report the same, 5.6 series may have been cut off, which OpenAI likes to do without ceremony nor announcement in the past.
Thank you. I went in and tried it. It works! However, A model selector hidden behind a tiny caret inside another control is easy to miss even with normal vision. For users with low vision, reduced contrast sensitivity, motor difficulties, or anyone using magnification, it can become effectively undiscoverable. And because “Medium” sounds like reasoning effort, not model selection, there’s no obvious semantic clue that a second control is hiding there.
It also needs to be documented plainly somewhere visible. Something as simple as:
To choose your Chat model: open the thinking-effort control (for example, Medium), then use the caret beside it to select GPT-5.6 or GPT-6.
And separately, Settings should explain whether Chat has a default model preference at all. Right now the interface gives users no clear indication that the model choice is per-conversation rather than account-wide.
That is exactly the kind of issue that gets dismissed as “just UX,” when in practice it affects accessibility, predictability, and user control.
This issue was solved in the communication above. It was raised solely because of poor discoverability in the ChatGPT user interface for a model family switch.
It is not quite obvious that the only way to see different models is to click here on the word “medium”.
