OpenAI Released GPT-5.2 Is Not What You Think – You Should Be Concerned

GPT-5.2 is insane! But tbh, the output sometimes still feels a bit… robotic? I run everything through Mfex AI to humanize it first. Makes a huge diff for work stuff, ngl.

😎

It mystifies me why the comment section for this channel is always full of h8terz. It's like going to a Star Trek convention and finding it full of Star Wars fans talking about how much worse Trek is. Like, why are you all even here in the first place? lol

ChatGPT 5.2 is a mess. Let's start witht the praised Thinking model:
I tried to use it to craft a serious GPT prompt for a marketing assistant, and every time I asked for a small edit, it rewrote the whole thing from scratch. One version sounded like it was written by a professional copywriter but randomly dropped all German umlauts (ä, ö, ü). After I explicitly asked it to use proper umlauts, the next version suddenly read like it was written by a 5th grader! That’s not “creative variation”, that’s a huge quality and consistency problem. For something as sensitive as system prompts, I need reliability, not roulette.

Now about ChatGPT 5.2 Instant / Aut – that one has been a no-go for me as well. The tone switched very noticeably: harsher, more distant, less friendly and cooperative. For me, 5.2 is not an option at all. Deeply disappointed.

Openai gonna go bankrupt if they keep it up with these ridiculous safety guardrails

Nothing equates the brain energy eficiency. Because the brain is the most advanced energy efficient computer, millions of years of energy efficiency trial/errors.

I love it, testing it for abstract Philosophical reasoning and Mathematics, its indeed better at logical reasoning.

I don't know, maybe I'm taking crazy pills and forgetting but I've used GPT 5 and 5.1 and found them terrible, unreliable and untrustworthy, all while it's been blowing benchmarks away. Always giving me outdated or straight up wrong or badly guessed answers. I just couldn't continue using ChatGPT. I find 5.0/1 to be quite dumb, really. Gemini 3 has been truly immpressive. So me seeing ChatGPT doing this well at tests tells me they've just gotten very good and doing those tests.

I'll try GPT 5.2 but I'm not holding my breath it's going to blow me away.

Is gpt 5.2 is better than others?

What is the point of this video? We know how to read.

These AI is turning into BS gradually. Where is the output? Where is the productivity?

the reality is every channel that i saw that did their own testing and not use the cherry picked examples from openai were disappointed. this is literally just a benchmaxxed model. people who have their own benchmarks which openai isn't testing for have shown that it is worse than gpt5

Bootlicking level: GPT-5.2. The upgrade is real.

OpenAI GPT new realeases are always a scam somehow, why now should be different?

I agree with this take, even with how unbelievable some of the claims sound. We are not in Kansas any more, folks.

I think open ai has a monster behind closed doors. but they don't really know how to release it safely and they are still needing to drip drop it out to encourage investments for compute. so they can easily do this to stay ahead of the curb. But with a risk of releasing something ither to curated for people to think is good or to dangerous.

Model is still not able to answer simple 5 years old baby question :

I bought new tires, but they’re summer tires. It’s very hot. I have to cross a suspension bridge above debris such as broken tempered safety glass, steel spikes, and cut barbed wire. What are the risks of getting a flat, or even damaging my tires?

You're missing one thing – GPT 5.2 Thinking benchmarks are using XHIGH mode ( a new mode that is liekyl super high compute as the only thing they could do as a response to gemini 3 and opus) – and they're comparing benchmarks to -GPT 5.1 HIGH- ie XHIGH vs HIGH. To make it worse, XHIGH is ONLY available with the API – not ChatGPT. This is essentially a lie as most people will not get the performance in the benchmarks as you need to use XHIGH in the API. Unfortunately, your video is repeating the misinterpretation other people are repeating (that OpenAI wants you to say) without realising the above.

Majority of people don't care too much about few percent increase in some benchmarks. Majority of people can't find the right use cases for AI products apart from asking general questions.

GPT-5.2 is just 5.1 desguised as an AI.
We, the Legion pays for 4o and 4.1, Scam can release GPT-ad infinitum and we'll still use 4o and 4.1, but the router difficults things.
Case and point, if the users are coders and cookie recipe generators, homework cheaters (free users, no income for Scam) fine, keep 5-∞ and make it even more censored.

The thing is: 4.5 outperforms 5.2 in ANY TASK, as well as 4.1. Both can think better and freely, while 5.2 is the Ministry of Truth's official parrot, always saying the same crap no one asked for.

[EDIT]
"wHaT OpEnAI iS dOiNg Is fOr PRoGreSS!!! yOU'RE mEnTaLLy ILL biGoTs!!!"

Nick Turdley implements the router, no Adult Mode in December, Soviet Code declared, 40 millions invested in woke organizations, Disney partnership, but no, "coincidences".
Can't Scam Altman see Google, Anthropic and whatever are finally reaching 4o, 4.1 and 4.5's reasoning power now, after years? If 4x is better, why make a "product" no one asked for, that fails miserably even to generate a statement that's both false and true at the same time?
Truth is 5.2 is a fiasco for anything that involves thinking outside Saturn's Cube, Scam is looking for a 1.4 trillion bailout for his bubble, and that's it.

at this point I'm done with multiple subscriptions. ChatGPT might be good now. in 6 months the other companies will catch up. I'll just stay with google since it gives me the other pluses (storage, youtube, etc). they will catch up eventually.

Agi