Sidebar Sidebar
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

ChatGPT Switches Default Model to GPT-4.1 Mini for All Users, Subscribers Get GPT-4.1

MacRumors

MacRumors

macrumors bot
Original poster
Apr 12, 2001
68,083
38,825


OpenAI has officially integrated its flagship GPT-4.1 model into ChatGPT, making it available to Plus, Pro, and Team subscribers starting today. The company announced the rollout on X (Twitter), highlighting the model's strengths in coding tasks and instruction following.

chatgpt-logo.jpg

Enterprise and Education users won't be left behind for long, with access promised "in the coming weeks," according to OpenAI.

Alongside the premium-only upgrades, the company is also replacing GPT-4o mini with GPT-4.1 mini as the default model for all users, including those on free accounts. If you've enabled the ChatGPT extension for Apple Intelligence on your devices, Siri will presumably now default to GPT-4.1 mini going forward.

Both models support a one million context token window (the amount of text/media in a prompt that an AI model can process), far surpassing GPT-4o's 128,000-token limit. OpenAI claims that speed improvements make GPT-4.1 "more appealing for everyday coding tasks" compared to their recently introduced reasoning models.
The deployment comes a few weeks after the new models were released to developers. OpenAI said that both new models have undergone standard safety evaluations before being released to the public.

Article Link: ChatGPT Switches Default Model to GPT-4.1 Mini for All Users, Subscribers Get GPT-4.1
 
Article Link
https://www.macrumors.com/2025/05/15/chatgpt-default-model-gpt-4-1-mini-all-users/
atownguy said:
I thought the default model has been 4o (not 4o mini)? And I think the default is still 4o- it just downgrades free users to mini once they hit their limits.
Click to expand...
For free members the default is also GPT-4o but the quota is very limited. It switched itself to GPT-4o-mini very quickly.

As a Plus member, I can confirm GPT-4.1 and GPT-4.1-mini are both available now, and GPT-4o-mini is gone.
 
  • Like
Reactions: arkitect
kiranmk2 said:
Soon, they will need a GPT model just to analyse your query and tell you which GPY model you should use!
Click to expand...
Seriously. It feels like they need to do a hard reset on the naming/versioning. I have a plus subscription, and it's still defaulting to 4o, with o3, o4-mini, and o4-mini-high as the main alternate choices. You have to go under "More Models" to find 4.1, 4.1-mini, and 4.5. I would love to assume that 4.5 is the latest greatest new hotness, but honestly I wouldn't put $100 on that bet. Is 4o better than o3? is o less than 1? or is it 15, since it's the letter "o"? why would I choose 4.1-mini over 4.1? Why would I use anything except 4.5? Does 4.5 have the same "advanced reasoning" as o3? 4.1 is for "quick coding", but o4-mini-high is "great at coding"... so I need to choose between quick and great code? Is there a model that does great quick code?

And biggest question of all, does any of this matter? If the latest version is more or less all the improvements of the previous versions all wrapped up in one, then why give all this choice? Conversely, if all these choices are all meaningful, then why is there virtually no guidance on when to use which one?
 
  • Like
  • Haha
Reactions: gusmula, It’s always something, Reverend Billy and 9 others
So I guess the plan is to just keep making free worse and worse until you have no choice but to start paying.
 
  • Like
Reactions: KeithBN
germanbeer007 said:
I stopped using chatGPT. Claude (windsurf, permanent early adopter pricing) + grok (x premium) suit my needs better and so far it's cheaper than paying for a single ChatGPT monthly license.
Click to expand...
Grok is great if you like getting answers unrelated to questions such as whether or not Toronto Blue Jays pitcher Max Scherzer is getting paid tens of millions for not playing
 
  • Like
Reactions: KeithBN
sw1tcher said:
Grok is great if you like getting answers unrelated to questions such as whether or not Toronto Blue Jays pitcher Max Scherzer is getting paid tens of millions for not playing
Click to expand...
That's my experience with ChatGPT so far actually. Grok 3 has been miles better than whatever OpenAI has been putting out.
 
I am shifting away from using ChatGPT. It has been making sill mistakes lately and I find Google and Claude better.
 
The naming and descriptions confuse the hell out of me.

I wish Siri queries saved in ChatGPT had their own folder or nested section (almost like what ChatGPT's done by adding an image gallery. Things are getting cluttered fast.
 
  • Like
Reactions: TracesOfArsenic and Dj64Mk7
jedivulcan said:
The naming and descriptions confuse the hell out of me.

I wish Siri queries saved in ChatGPT had their own folder or nested section (almost like what ChatGPT's done by adding an image gallery. Things are getting cluttered fast.
Click to expand...
Funny you mentioned the obsolete and useless Siri





I don’t use that anymore 🤣🤣


My new confidant is ChatGPT.

It pulled me out of depression and gave me some sound health related advices.

Tried and tested

I’m living proof

So no one can tell me otherwise!!!
 
  • Love
  • Wow
Reactions: gusmula and klm9210
AgeOfSpiracles said:
Seriously. It feels like they need to do a hard reset on the naming/versioning. I have a plus subscription, and it's still defaulting to 4o, with o3, o4-mini, and o4-mini-high as the main alternate choices. You have to go under "More Models" to find 4.1, 4.1-mini, and 4.5. I would love to assume that 4.5 is the latest greatest new hotness, but honestly I wouldn't put $100 on that bet. Is 4o better than o3? is o less than 1? or is it 15, since it's the letter "o"? why would I choose 4.1-mini over 4.1? Why would I use anything except 4.5? Does 4.5 have the same "advanced reasoning" as o3? 4.1 is for "quick coding", but o4-mini-high is "great at coding"... so I need to choose between quick and great code? Is there a model that does great quick code?

And biggest question of all, does any of this matter? If the latest version is more or less all the improvements of the previous versions all wrapped up in one, then why give all this choice? Conversely, if all these choices are all meaningful, then why is there virtually no guidance on when to use which one?
Click to expand...
I literally asked it which model to use for my needs and recommended the pros and cons... So I feel you can just do this to answer your comments (spoiler alert: 4.1 was the best and more accurate, especially with the newly expanded token limit).
 
I’m so surprised by the number of people who are happy with ChatGPT. These past few weeks, I don’t think I’ve had a single correct answer from it, even for the simplest questions. It doesn’t take my settings into account, it hallucinates answers, it makes completely false statements, and it’s been totally unusable to the point that I cancelled my subscription. I had to ask it to check its answers every time, and every time I would get “You’re right, I made a mistake,” literally every single time. Even the translations were completely off (I’m French, I understand English perfectly but I’m not very good with grammar). Once it even translated into German....
 
  • Like
  • Love
Reactions: Will Tisdale 🎗, surfbutler, KeithBN and 1 other person
bobbyjack said:
I’m so surprised by the number of people who are happy with ChatGPT. These past few weeks, I don’t think I’ve had a single correct answer from it, even for the simplest questions. It doesn’t take my settings into account, it hallucinates answers, it makes completely false statements, and it’s been totally unusable to the point that I cancelled my subscription. I had to ask it to check its answers every time, and every time I would get “You’re right, I made a mistake,” literally every single time. Even the translations were completely off (I’m French, I understand English perfectly but I’m not very good with grammar). Once it even translated into German....
Click to expand...
Paid or free version?
 
Paid :/

I just tried 4.1 and I still get false results. No, ChatGPT, Uma Thurman isn’t 36 anymore but I agree she looks great for her age.. I opened a new window and asked the exact same question, and this time I got the correct age, it's completely random.

I’ve been lurking on the ChatGPT subreddit and it seems it's a recent issue.

Edit : I just uploaded a 4 pages pdf, asked him to analyze it and got "Your file “XXXX.pdf” is completely blank and contains no readable text or usable information. There is no data to translate in this document."

I told him it was false, and then I got "I will analyze the file content again and extract the text precisely to translate it into French. Please wait during the extraction. Upon verification, the file consists only of scanned images, with no selectable text. I will extract the text from the images, then proceed with the translation into French. Please wait during the processing.".

Mind you, I wasn’t even asking for a translation, I wanted an analysis of the document...
 
Last edited:
  • Like
Reactions: Will Tisdale 🎗 and KeithBN
bobbyjack said:
I’m so surprised by the number of people who are happy with ChatGPT. These past few weeks, I don’t think I’ve had a single correct answer from it, even for the simplest questions. It doesn’t take my settings into account, it hallucinates answers, it makes completely false statements, and it’s been totally unusable to the point that I cancelled my subscription. I had to ask it to check its answers every time, and every time I would get “You’re right, I made a mistake,” literally every single time. Even the translations were completely off (I’m French, I understand English perfectly but I’m not very good with grammar). Once it even translated into German....
Click to expand...
I told you not to ask how to turn lead into gold nor should you ask how to remain immortal.

Tut tut

You never listen.
 
You must log in or register to reply here.
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.
Top Bottom
Back