ChatGPT pushed out yet another set of updates, shuffled the model menu, and renamed a few things for reasons known only to OpenAI’s internal naming committee.
Lucky for you, I read the documentation so you don’t have to.
If you prefer the video, watch here:
If you log in today, you’ll notice the default model is now GPT 5.1 instead of GPT 5.
You still get the same three major modes:
Think of it like “casual drive,” “sport mode,” and “off-road with a roof rack full of logic.”
When GPT-5 first came out, many people said it felt colder and more robotic than GPT-4. The emotional tone dropped off a cliff.
GPT 5.1 Instant is OpenAI’s attempt to climb back up that cliff.
According to the docs, it’s now:
Translation: conversations should feel more human” again, and less like you’re arguing with a politely annoyed spreadsheet.
OpenAI calls it their “advanced reasoning model.”
Their claims:
My experience so far? It’s… fine. Not mind-blowing, not disappointing. Just fine. You’ll need to try it on your own use cases to get a feel for it.
OpenAI managed to take something simple and give it the IKEA treatment.
Here are the numbers in plain English:
For GPT 5.1 Instant
For GPT 5.1 Thinking
And if Auto switches models for you, that usage doesn’t count toward your limits at all. Make of that what you will.
If you’re wondering how many tokens you get before ChatGPT starts forgetting everything you said like it’s rebooting an Etch A Sketch:
GPT 5.1 Instant
GPT 5.1 Thinking
Compared to competitors:
So 196K isn’t wild, but it’s better than before, and most users won’t hit that ceiling unless they’re feeding it a mid-sized codebase.
If you click your menu (bottom left as of today, but OpenAI likes to move things around), you’ll see updated names for tone settings.
Notable changes:
These seem like label changes, not behavior changes.
The docs mention you can now adjust “how concise, warm, or scannable results are” and control emoji usage, but that option doesn’t actually exist in the UI yet. Maybe soon. Maybe never. Hard to say.
A few minor changes from the step-by-step presented in this article I published a few days ago.
If you miss the GPT-5 era — I won’t judge — you can still manually switch back to:
Free users can’t choose models; paid users can.
GPT 5.1 isn’t revolutionary ( if it were, they’d call it 5.5 or 6.0 ) but it’s a meaningful update:
Try Instant and Thinking on a few of your real tasks and see what fits. Most people will notice small quality-of-life improvements more than big breakthroughs.
Let us know how it works for you.