There is a slightly confusing moment in ChatGPT’s model picker.
Instant, Medium, High and Extra High all look like different amounts of reasoning applied to GPT-5.6 Sol.
Then you move to Pro and the screen may suddenly say GPT-6 Pro.
That raises the obvious question:
“Is Pro just Sol thinking harder, or did ChatGPT actually switch me to Astra?”
And the important detail is that we are talking about normal Chat, not Work.
As of September 7, 2026, OpenAI’s official documentation gives a clear answer: Extra High is still GPT-5.6 Sol, while GPT-6 Pro is powered by GPT-6 Astra.[1]
So if the UI explicitly changes to GPT-6 Pro, this is not merely “more reasoning on Sol.”
The model family itself has changed to the GPT-6 Astra side.
That is the moment when “okay, paid plans are doing more than I thought” becomes a reasonable reaction.
1. The simplest table — Extra High and Pro are not just two adjacent steps on one ladder
For normal Chat, OpenAI currently describes the options like this.[1]
| Option | Main model | What changes? |
|---|---|---|
| Instant | GPT-5.6 Sol on eligible paid plans | Speed-first; may automatically reason more on complex requests |
| Medium | GPT-5.6 Sol | Standard reasoning effort |
| High | GPT-5.6 Sol | Longer reasoning |
| Extra High | GPT-5.6 Sol | Highest reasoning effort available with Sol |
| Pro | GPT-5.6 Sol Pro or GPT-6 Pro | Uses a Pro model, not merely another Sol effort setting |
| GPT-6 Pro | GPT-6 Astra | The model family changes to GPT-6 |
The key distinction is simple.
Extra High means “how hard should Sol think?”
Pro is not automatically “Extra High plus one more notch.”
OpenAI explicitly says the Pro models can include GPT-5.6 Sol Pro and GPT-6 Pro.[1]
So the UI may look like one continuous reasoning scale, but the last door can lead to a different model entirely.
2. Up through Extra High, it really is GPT-5.6 Sol
This part is straightforward.
OpenAI says that, on eligible paid ChatGPT plans, Instant, Medium, High and Extra High are powered by GPT-5.6 Sol.[1]
Extra High is specifically described as the highest reasoning level available with GPT-5.6 Sol.[1]
That makes the sequence
Medium → High → Extra High
best understood as changing the amount of reasoning available to the same Sol model, rather than swapping to another model family.
So far, so expected.
The surprise starts at Pro.
3. Pro is not “Sol Extra Extra High”
The word Pro makes it tempting to assume:
“This must be the final form of Sol, just with even more thinking time.”
But the product structure is different.
There is a GPT-5.6 Sol Pro model, and there is also the newly rolling-out GPT-6 Pro.[1]
That means Pro is not only a reasoning-effort label. It is also a lane for higher-end Pro models.
If your selection is effectively
GPT-5.6 Sol → Pro,
you may still be on the Sol family through Sol Pro.
But if the UI explicitly says
GPT-6 Pro,
that is a different story.
That name points to Astra.
4. GPT-6 Pro is powered by GPT-6 Astra
OpenAI says this directly.
Its Help Center currently describes GPT-6 Pro as powered by GPT-6 Astra and says it is rolling out in ChatGPT.[1]
OpenAI also announced GPT-6 Astra in its September 3, 2026 release notes, describing it as its most capable model for difficult end-to-end work, with improvements in coding, research, computer use and complex multi-step tasks.[2]
So the clean mental model is:
GPT-6 Pro = the ChatGPT Pro form that uses GPT-6 Astra.
That resolves the original confusion.
Extra High is still Sol.
Once the UI says GPT-6 Pro, you are no longer in “Sol but trying harder” territory.
You are in Astra territory.
5. “But I mean Chat, not Work” — OpenAI treats those rollout surfaces separately
This distinction matters because Astra is not rolling out to Chat, Work and Codex at exactly the same time for everyone.
OpenAI explicitly says availability can differ between the three products during rollout.[1]
As of September 7, 2026, GPT-6 Pro is rolling out in normal Chat for Pro $100, Pro $200, Business and Enterprise plans.[1]
For Plus, Astra is being rolled out in Work and Codex rather than normal Chat.[1]
That means saying “I can use Astra” is not enough to tell you where you can use it.
Someone may have Astra in Work before they have GPT-6 Pro in Chat, or the reverse for a different eligible plan.
So the clarification “not Work — I mean normal Chat” is not nitpicking. It is exactly the right product-level question.
6. The UI changed when you switched — then the model probably really changed
This was the fun part of the whole discovery.
First, you assume Extra High is Sol.
Then you switch to Pro.
The UI changes to GPT-6 Pro.
You ask, “Wait, is this Astra?”
You check the official documentation.
Yes. It really is Astra.[1]
The practical lesson is to look not only at the word “Pro,” but at the actual model name shown after the switch.
- GPT-5.6 Sol Pro means the Sol family.
- GPT-6 Pro means the Astra-backed GPT-6 family.[1]
Same broad Pro area, different model lineage.
7. Then the model’s own self-introduction started wobbling — trust the badge more than the introduction
There was one more comic layer.
When asked “what model are you?”, the answer bounced between Sol and Astra and even included a correction.
In other words, asking the model to verify its own identity produced a slightly unstable self-introduction.
For this case, the more reproducible evidence was:
- What the app’s model picker actually displays.
- How OpenAI’s current official Help Center defines that displayed model name.
Here the UI said “GPT-6 Pro,” and OpenAI says GPT-6 Pro is powered by GPT-6 Astra.[1]
That is enough to connect the dots.
The blunt practical rule is:
When the AI’s self-introduction and the product UI disagree, check the name badge and the official registry.
8. Paid plan power, yes — but Pro does not mean unlimited GPT-6 Pro
The reaction “wow, that is a serious paid-plan feature” makes sense.
Normal Chat can now expose a different top-end model family rather than simply giving Sol more reasoning effort.
But OpenAI also states that GPT-6 Pro has usage limits.[1]
A Pro subscription therefore does not mean unlimited GPT-6 Pro usage.
And because rollout is gradual, Chat, Work and Codex may still light up at different times on the same account.[1]
Early September 2026 is very much a period where the product matrix is changing in real time.
9. Summary — Extra High stays on Sol; if it says GPT-6 Pro, it is Astra
The whole answer fits into six lines:
- Instant / Medium / High / Extra High: GPT-5.6 Sol on eligible paid ChatGPT plans.[1]
- Extra High is the maximum reasoning effort for Sol.[1]
- Pro can use GPT-5.6 Sol Pro or GPT-6 Pro.[1]
- GPT-6 Pro is powered by GPT-6 Astra.[1]
- Chat, Work and Codex can roll out at different times.[1]
- If the post-switch UI explicitly says GPT-6 Pro, the model has moved from the Sol side to the Astra side.
Shortest version:
Extra High asks Sol to think harder. Switching to GPT-6 Pro can change the player entirely: Astra comes in.
So yes — the Pro switch really can change the model, not just the reasoning level.
Sources
- OpenAI Help Center — GPT-5.6 and GPT-6 Pro in ChatGPT help.openai.com
- OpenAI Release Notes — Introducing GPT-6 Astra (2026-09-03) openai.com
