fresh bayReported by @cunning fox
#GPT 5 models abusing thinking tools even when not being told or using the tool
cunning fox
cunning fox
it is annoying
vital tapir
cunning fox
Now i can only talk for 3 times it gone normal...then always abuse thinking tools
kind niche
Now it's doing it to me with every chat unless I am on 4.1 x.x
cunning fox
Now it's doing it to me with every chat unless I am on 4.1 x.x
Can u skip the thinking or receive same errors as mine?
Nice chatgpt is getting bugger throu days
kind niche
Mine will not let me skip and while I don't get errors, it will not stop defaulting to thinking. In any chat unless I am using 4.1
swift root
Same problem here, it is triggering the Thinking all the time
kind niche
Its frustrating cause I have really began to enjoy Auto 5 over 4o and 4.1 for creative writing. It seems to be more fluid. And way less repetitive with the "You're not alone, you're never alone," crap when I am writing heavy scenes or fight scenes with multiple characters. Smh.
cunning fox
Mine will not let me skip and while I don't get errors, it will not stop default...
Same and its getting worse, every thing abit complex it wil sctivate the bot
viscid pebble
I have this issue in projects!!! Business plan. Thinks aprox 4 to 5 min even if set on fast thinking
also now and then node errors. Or If I paste info it vanishes
hearty fable
Plus-user. Yesterday my SVG-question triggered Thinking and after that I have had only 3 responses (different chats) in Auto, otherwise it has been Thinking-mini. Unless I choose 5-Instant or 4.1.
What is happening with 5? I don't need reasoning in every response.
viscid pebble
Plus-user. Yesterday my SVG-question triggered Thinking and after that I have ha...
I tried projects last night over 1 minyute thinking on Business account... on any model!!!
cunning fox
why is this already almost 1 month but openai still not fixing this issue?
i want the thinking model disappear not reduce thinking time...:'(
visual plume
try "quickly answer" in your prompt.
Otherwise, you can set param reasoning:{'effort':'minimal' or 'low'} when post the request if you use sdk
I fixed most of problems in my project ：）
cunning fox
it still the same!
cunning fox
I fixed most of problems in my project ：）
for both normal model, a variants i ffound on gpts...and the project...nothing changed at all!
always abusing thinking modelll..no matter what promts i used
visual plume
I use gpt5-mini and gpt5.
Do you mind to provide a case of your prompt?
Maybe I can test it for you ?
cunning fox
first i tried round brackets like (dont abuse thinking tools)
second time i tried stop its rsponse and retry to enable the skip button but it said it cant found the conversation in some kind of tree
the third time i try is ur quick answer promt...and the last then i use whatever said dont use or dont abuse thinking tools or any tools, etc.
visual plume
wired 🙁
cunning fox
😑
viscid pebble
guys, IMHO that is not thinking as you think... but mostly one of the moderation steps on the output the ai generates
viscid pebble
🤨
sounds wicked? heh... if input is fine and you have long waiting for output the ball is in openai side
cunning fox
sounds wicked? heh... if input is fine and you have long waiting for output the...
i didnt just got long waiting but i also got the model abusing the thinking tools no matter how i tries...and now it dont even shows the skip button, and not even let me refresh the response!
😑
cunning fox
chhhhhh.... windows app, browser or android?
phone app(both apple and samsung), google, uh..firefox...microsoft edge...and many manyother tries...still thinking tools ultra abusetion!
😠
😡
nothing works!
and now it is making the whole website goin into a lag situation!
dire zinc
Same here guys. If I say just hello, it’s start thinking for better answer and can’t skip it. Than it gives me a very long bad answer. If I choose long thinking mode than I can skip it and I get the normal answer. But if I’m waiting suddenly the message delete itself or it give me the long thinking answer
cunning fox
Same here guys. If I say just hello, it’s start thinking for better answer and c...
lol even for hello?! XD
what the hell? XD
dire zinc
Yes it doesn’t matter what I ask 😂 and if I say stop thinking and answer immediately, it’s thinking again and it is giving me a “short quick answer”
cunning fox
same here bro
the thinking model is stupid or something...a simple quiestions of math.,,it cant answer immediately...and when thinking done...it gives wrong answer...
dire zinc
I need to use all my tokens for the best gpt 5 model ever and then I get the worse gpt 4 model which responds better lol. this is so stupid
A week ago this never happened and everything was fine
cunning fox
A week ago this never happened and everything was fine
yea true
hearty fable
lol even for hello?! XD
Yesterday was pretty ok, couple of thinking-mini responses, then good hours in Auto and then when I said "Let's continue tomorrow. Good night" -> It thought 15 seconds just to responds good night. Like what .. 🤣
cunning fox
Yesterday was pretty ok, couple of thinking-mini responses, then good hours in A...
Ain't no way...i have to test that
random stone
Yesterday was pretty ok, couple of thinking-mini responses, then good hours in A...
Yes!! It’s thinking longer to say “bye” than a human brain does to do basic math. I thought this thing was supposed to be super intelligent?
hearty fable
But it does not show reasoning to me. Would be interesting to see what it is thinking 😅 "User seems to want to go to sleep. Maybe it is appropriate to say Good night to user"
cunning fox
How can i out of free thinking model and the model still can hop in?!
Why do it say i am reached limit while i still using normal model like nothing?
hearty fable
Why do it say i am reached limit while i still using normal model like nothing?
If the Router (Auto) triggers Thinking modes, they are not counted as limited use. Only if yourself select Thinking or T-mini, they are counted.
cunning fox
If the Router (Auto) triggers Thinking modes, they are not counted as limited us...
How do i disable it?
dire zinc
Doesn't everyone have this problem?
quiet bane
yes
hearty fable
How do i disable it?
You don't, it is build in coding for LLM.
But it seems that using legacy models 4o and 4.1 don't trigger Thinking modes. From 5-series you can use 5-Instant to get rid of thinking.
cunning fox
You don't, it is build in coding for LLM.
But it seems that using legacy models ...
No matter how i tries...no matter what i do...it stillk abuse thinking
viscid pebble
No matter how i tries...no matter what i do...it stillk abuse thinking
if you note this means you are restrained because of moderation flags.
move any "co-writing" or RP to third person, avoid prohibited topics.
After a couple of days it will relax. But not garanteed
umbral hornet
You don't, it is build in coding for LLM.
But it seems that using legacy models ...
this seems to not be true
even when selecting instant it is using thinking and the moment it does it absolutely destroys the chain of conversation
hearty fable
I have now had 2 good days with Auto not routing. Today I mentioned before Sora 2 live that maybe I should get a beer before live starts and boom -> thinking-mini. I even rousted him about that am I now on sensitive topic cause I mentioned beer 🤣
cunning fox
i wonder when will openai stop this madness
viscid pebble
Routing is all normal. Context understanding has been restored.
Any complaint? Verify if your exchanges are in accordance to policies. IMHO they are not
rich wigeon
I am using gpt4o and it suddenly start to use thinking for every answer. and the answer i get is gpt5 style. I hate gpt5 style which is the reason i am using 4o, and now this??