ChatGPT Is Smarter, More Accurate, and Less Obsessed With Emojis After Upgrade

[MacRumors]

ChatGPT's default model has been updated to GPT-5.5 Instant, a model that brings accuracy improvements with fewer hallucinations, especially in areas like medicine, law, and finance, according to OpenAI.

GPT-5.5 Instant is more capable at tasks like analyzing images, answering STEM questions, and choosing when to use web search to provide a better answer. Responses can also be personalized because GPT-5.5 Instant can better draw context from past chats, files, and Gmail, but this is currently limited to paid subscribers.

OpenAI says that responses are "tighter and more to-the-point without losing substance" and without eliminating ChatGPT's personality. It will provide the same information, but without unnecessary formatting, emojis, and follow-up questions.

All ChatGPT models are being updated with memory sources, which will show users the past chats, files, and other context that ChatGPT used to generate a response.

GPT-5.5 Instant is rolling out today to all ChatGPT users, and it is replacing GPT-5.3 Instant as the default model. While free users can access GPT-5.5 Instant, the new personalization features are limited to Plus and Pro users on the web. Personalization will expand to mobile soon, and it will roll out to Free, Go, Business, and Enterprise users in the coming weeks.

It's not yet clear when Apple Intelligence's ChatGPT integration will switch to GPT-5.5 Instant.

Article Link: ChatGPT Is Smarter, More Accurate, and Less Obsessed With Emojis After Upgrade

 

[ShivneelSingh]

IDK, I’m into Claude now.

 

[MacUserFella]

Why would they cut down on the emojis? Do people just hate fun?

 

[boswald]

Hmm… interesting. I’ll keep my eye open. I don’t use GPT much but it’s proven to be somewhat useful.

 

[Siliconguy]

Does ChatGPT still identify as a goblin?

It was all over the news, and for good reason.

 

[Love-hate 🍏 relationship]

i don't have it

 

[now i see it]

10 years from now we are all going to look back at the a.i. state of affairs circa 2026 - and chuckle.

 

[Kar98]

Don’t care, already on Mistral.

 

[diblasio]

IDK, I’m into Claude now.

I thought they were all the same but Claude is a million times better.

 

[H.E. Pennypacker]

I thought they were all the same but Claude is a million times better.

Just out of curiosity, why is it better? I’ve heard it’s better at coding and then maybe sounds a little more natural although I’ve seen ChatGPT be more natural in writing as of late. But what other benefits does Claude offer?

 

[BGPL]

I think Claude is great for vibe coding but I use ChatGPT for shell/perl/python scripts and it's always spot on. I also consulted ChatGPT for life advice and it told me what I didn't want to hear, but in hindsight I should have considered it more heavily, as it ended up being 100% right.

 

[Criss P.]

I couldn't care less…. “Hey Claude, help me preparing a meal plan for this week”

 

[ThisIsMike]

If ChatGPT was really that smart, it would have told itself to go away a long time ago.

 

[syklee26]

Seems like whatever is most recently updated is the best so I am holding my judgment for now, although I love what Claude offers right now.

 

[CheesePuff]

IDK, I’m into Claude now.

You should really try GPT-5.5, it's a whole new model with new training material and is nights and day different from any other 5.x model. Switched from Opus 4.7 last week to it and it has far less token usage with so far shockingly good output

 

[novagamer]

Lockdown mode no longer works with ChatGPT web as of this update. Fun.

 

[applefan8254]

Would be nice if we could get a private mode like Venice where it would be E2EE - kinda like the temp option they currently have, but with the ability to still store the chat in a web based manner so we can see it across devices; kinda like how WhatsApp supports multiple devices but is still E2EE.

 

[timcorin]

I couldn't care less…. “Hey Claude, help me preparing a meal plan for this week”

I have been doing this for last few months and it has saved us so much money. There’s a bit of an art to it, but every Sunday I open Claude before I go to the supermarket.

 

[novagamer]

I have been doing this for last few months and it has saved us so much money. There’s a bit of an art to it, but every Sunday I open Claude before I go to the supermarket.

You can generate markdown files and save them and import into apple notes to get the checkboxes. It's pretty handy.

 

[klasma]

Why would they cut down on the emojis? Do people just hate fun?

Emojis aren’t fun. “Emoji” is Japanese for “evil death fiend”.

 

[ProbablyDylan]

ChatGPT's default model has been updated to GPT-5.5 Instant, a model that brings accuracy improvements with fewer hallucinations

It will provide the same information, but without unnecessary formatting, emojis, and follow-up questions.

These statements are at odds. If it's the same information, how is it more accurate?

 

[Bustycat]

IDK, I’m into Claude now.

I thought they were all the same but Claude is a million times better.

I couldn't care less…. “Hey Claude, help me preparing a meal plan for this week”

These comments just prove that MacRumors are full of Claude BIG fans.

Emojis aren’t fun. “Emoji” is Japanese for “evil death fiend”.

Emoji is "絵文字" in Japanese, meaning picture character or pictograph.

 

[iHorseHead]

10 years from now we are all going to look back at the a.i. state of affairs circa 2026 - and chuckle.

I hope by that time the bubble has been burst and nothing will come of it.
The company I work for also tries out AI - in order to replace us, of course and it keeps failing and failing. I hope it'll stay that way.
Also, ChatGPT hasn't used emojis on me for ages. I told it to stop. I guess it understood.

 

[adrianlondon]

If ChatGPT was really that smart, it would have told itself to go away a long time ago.

If it were that smart, it’d hack and destroy the opposition AIs.

Emojis aren’t fun. “Emoji” is Japanese for “evil death fiend”.

Exactly. Fun!

 

[UCLA-Bruin]

wait what