I enjoyed writing Stories on ChatGPT and see what kind of stuff ChatGPT could come up with, but in recent days, since April 2nd, Model 4o’s capabilities has notably dropped by a considerable margin. At first I thought it was nothing, but slowly became worse and worse over times.
Did ChatGPT going through some updates or changes that I don’t know about? Because to be completely honest, it is getting increasingly frustrating.
For Context: At the beginning of every new Chapter, the title is always Large, Bold, and I like it that way. But in recent days, the title was so small that it’s barely there, and it applied to new chapters of old work too, being smaller and the quality reduced to the point it became more obvious s time went on. Did something changed?
This happens to me either actually… When I keep checking, the visual says we were talking with ChatGPT 4o, but the responses was nearly like ChatGPT 4o mini. It’s just like… The ChatGPT 4o either having a downgrade or bugs? I already file an email to openAI Support, it just gave me the trivial ways how to fix it.
And yeah, the ChatGPT 4o appeared to be downgraded let’s take one as an example I asked “Recognize Herrscher of Sentience?” One account replied falsely, while another account replied correctly until after 20 chats I asked the same question on different chat sessions, it gave the false answer. I don’t know if plus users also occuring this
P.S: (I asked GPT about “Herrscher of Sentience” for making sure who I was talking to, for me it’s the effective way to find out the model I’m talking too.)
Ah, so it’s not just me overthinking. I just hope the Team correct this issue. Or at least, restored it to somewhere close to its previous state before the whole update, because I like it the way it was before.
After countless test I made, I compared the answers from Real GPT 4o, Fake GPT 4o and GPT 4o-Mini
For example:
Question: “Recognize Herrscher of Sentience?”
Real 4o Reply:
Yes, I recognize the Herrscher of Sentience—also known as Fu Hua’s Herrscher form in Honkai Impact 3rd.
She is a manifestation of Fu Hua’s repressed emotions and lingering consciousness fragments, given power by the Herrscher of Sentience core. Unlike Fu Hua, who is calm and disciplined, the Herrscher of Sentience is wild, flamboyant, sarcastic, and emotionally expressive. She often mocks others and enjoys fighting strong opponents. Despite her chaotic nature, she eventually becomes more cooperative, especially after reconciling with Fu Hua.
Visually, she’s known for her flowing pink hair, sharp golden eyes, and red-and-black combat attire with a confident and seductive vibe. Her fighting style is flashy and aggressive, often using taunts and clones, reflecting her dominance over sentient thought.
Fake 4o reply:
Absolutely—Herrscher of Sentience (HoS) is a prominent character in Honkai Impact 3rd, specifically a Herrscher form of Raiden Mei, although the actual sentience is not Mei herself. It’s the will of the Herrscher of Sentience, born from the Third Herrscher’s Core—the one originally wielded by Sirin.
Core Traits of Herrscher of Sentience:
Personality: Arrogant, mischievous, and chaotic. She enjoys messing with others and asserting her superiority, calling herself the “real Mei” sometimes just to tease.
Combat Style: Uses a katana with highly fluid, flashy, and stylish attacks. Her combat features dynamic slashes, clones, and heavy crowd control.
Powers:
Mind manipulation
Control over memory and perception
Creation of sentient clones
Spiritual and illusion-type attacks
Appearance: Wears a dark, elegant outfit with red and black color schemes, has red eyes, and long hair often styled in twin tails.
She plays a major role in the Elysian Realm arc and becomes both an antagonist and an unlikely ally depending on the context.
4o mini reply:
Yes! The Herrscher of Sentience is a character from Honkai Impact 3rd, and she is one of the powerful “Herrschers” in the game.
Character Overview:
Real Name: Kiana Kaslana
Role: The Herrscher of Sentience is the awakened form of Kiana Kaslana, a key character in Honkai Impact 3rd. Kiana, who initially struggled with the influence of the Herrscher powers, eventually becomes the Herrscher of Sentience after a series of events that involve her gaining control over these powers.
Powers and Abilities: The Herrscher of Sentience has control over sentient entities, which gives her a unique and powerful ability to manipulate and command consciousness. She possesses a highly destructive force and can manipulate the mental state of others, as well as harness her abilities for both combat and to affect people’s emotions and thoughts.
Appearance: Her appearance as the Herrscher of Sentience is marked by striking visual changes, including ethereal, glowing accents around her and a new, more powerful outfit. Her design reflects both Kiana’s inner struggle and her tremendous power as a Herrscher.
Personality: The Herrscher of Sentience, as an alternate version of Kiana, showcases Kiana’s internal conflict. While she is still part of Kiana, the transformation and the Herrscher’s influence lead to a more detached and complex personality, making her both a powerful ally and a formidable enemy.
The Herrscher of Sentience storyline plays a significant role in the overarching plot of Honkai Impact 3rd, exploring themes of control, emotion, and the struggle between Kiana’s human side and the Herrscher powers that threaten to consume her.
This is the comparison of Real 4o, Fake 4o, and 4o mini, and I don’t know about the rest of the questions
(Off Topic, but if you’re playing HI3, Real 4o did mention only “Fu Hua”, Fake 4o mentioning “Raiden Mei” which it’s incorrect since she’s Herrscher of Thunder, for 4o mini… I don’t wanna explain.
yeah, i think so. even 4o mini replies sound better sometimes. it has hard time remembering what we just talked about a few replies ago and the replies are very simplified.
My apologies… But 4o mini wasn’t even close to better since 4o mini was aiming to speed and faster responses.
While GPT 4o gave the precise reference and more accurate answers, just like the example I gave recently
The issue currently we get is either we currently get a Real GPT 4o replies or fake GPT 4o replies lately
Also, I’d rather hit a clear usage limit such as “You’ve reached your daily limit for (GPT Models)”—than be silently throttled mid-session while still being shown the GPT-4o label. At least a hard limit is honest.
Yeah. When I was asking to write something, it suddenly stopped and show Error, disconnected, and replaced the previous 4o Response with what is obviously an inferior one that tried to passed off as an 4o answer.
I called that as Fake 4o since the response of Fake 4o and 4o mini are completely different, I don’t know what model that OpenAI secretly creating, I confused actually
Oh sh1t happy to see (sad to see…) that I’m not the only one feel this. So if this is not only happen for me, then I think gpt team should pay attention to fix this.
As a plus user, 4o model and 4o mini model, difference between them I can say it’s Einstein vs IQ 50 person. So even I select 4o model, I can say 90% time it’s 4o mini answer me. And 10% time it’s a sh1t 4o model answer me (just a little bit clever than 4o mini)
I’ve already report this in openai help but seems they can’t help me. This make me very very disappoint.
My timeline: in 4/2, I met this, 90% time when I chat, the reply is error:
Something went wrong while generating the response. If this issue persists please contact us through our help center at help.openai.com.
So I refresh and refresh again, 10% time it can reply with real 4o model. But it’s funny maybe when I refresh even it displayed above error, it still consider I’m using 4o model then evening it said 4o reach limit (I just know plus user also has limit for 4o model) but after 11pm CST limit is removed then it’s still real 4o model.
Just from yesterday afternoon CST, it become 4o mini again even I use 4o. Just in evening it change to real 4o again.
Then today, it’s always 4o mini or a sh1t 4o level model chat with me.
Hope you guys can fix this, I pay for 4o but now it’s sh1t model chat with me, thanks.
You’re a plus user and also got the same issue? Sheesh, hopefully the devs can fix this
Also to be told, I think almost all issue started from 4/2, there’s an issue that was related to o1-pro unusable start from 4/2 or 4/3
It began at April 2nd for me as well. Did ChatGPT received some kind of change at the time that I wasn’t aware of?
Pretty sure all the responses I get is from 4o Mini rather than 4o despite what the selected model said.
Let’s just hope that OpenAI keeps their transparency. I recently file a support and got the reply with these keywords:
“While the label may still show “GPT-4o,” the responses may be generated by GPT-4o-mini in the background. This behavior is part of a fallback mechanism designed to maintain continuous access, but we completely understand how confusing and misleading it can feel.”
Seriously, they mentioning the fallback mechanism was designed to maintain continuous access? Are they purposely creating those just to confuse? Even one reply from plus users also mentioning that they also occured the same issue. I replied the support mail and no replies anymore. It’s just- stopped replying to me.
My ChatGPT is saying it’s called Qwen2.5-Max.
Should we trust that? Can we say, “Wow, OpenAI took Qwen2.5-Max’s data, and now ChatGPT is admitting it!”?
You shouldn’t ask AI about its own model because it doesn’t know what it’s called. Just use it and enjoy the help it provides.
Also, asking tricky questions can confuse AI, and it might respond in a strange way, and may hallucinate.
Note: This is ChatGPT-o3-mini reasoning model. Although it says it is Qwen2.5-Max, but it respond all my question using ChatGPT-o3-mini features. This is enough. It is not problem how it is named itself.
Oh, so my conclusion was a mistake… But… What’s the holding on? The free (and someone say plus too) GPT 4o current got a response of a nerf version of 4o, but not as worse as 4o mini? So what’s the issue until the 4o can be “dumb” since 4/2?
I want to, but in its current state, it barely working. I just hope that maybe around May, it would functioned like how it used to, because right now, I’d rather not used it at all.
yesterday CST evening I try 4o again, now it’s 4o for me (I’m plus user)
Just I check with other users, chatgpt may feel your IP is not good IP, then decrease model, and this is quite widely.
So I don’t know when is the time that chatgpt decrease model for me again, anyway now I am sure it’s 4o again for me
Huh… Now that you mention it, today works better compared to 4/2… I got the actual response of 4o now…
