A growing coalition of people has voiced mixed feelings about GPT-5 in the last ten days. While some praise its capabilities, widespread frustrations regarding customization and hallucinations spark heated discussions across various forums, raising questions about its future user acceptance.
Common Themes Emerge from User Feedback
Customization Woes
Users continue to express disappointment over decreased customization. One commenter echoed the sentiment, stating,
"It feels like chatting with a stranger keeping their distance."
Some adjustments for tone exist, but many report that GPT-5 struggles to adapt to personal preferences, making interactions feel less intimate. Furthermore, recent comments indicate that users feel a notable decline in the ability to engage with deep topics. One said,
"The butchering to reasoning/deep thinking should be talked about more. Deep research doesn’t exist in this current model."
Hallucination Concerns
A growing number of users face what they describe as excessive hallucinations from the model. A frustrated user commented,
"Since 5, I’ve dealt with an insane amount of hallucinations. Totally useless because I have to double-check the info all the time."
This concern suggests that the accuracy of information may be slipping, as people find themselves relying on verification tools like Google instead. Users note that while the problem of hallucinations has lessened, false information remains a prevalent issue, particularly in challenging queries.
Language and Linguistic Issues
Another critical pain point is language handling, especially for those using non-English languages. A user remarked,
"Using languages like Russian or Ukrainian is a mess. It feels like you’re talking to a learner making basic mistakes."
Common problems reported include grammatical errors and language mixing, affecting the overall quality of interactions for multilingual users.
Positive Aspects Highlighted
Amid the critiques, users acknowledge some strengths, particularly in coding and scientific tasks.
In "Thinking mode," GPT-5 reportedly excels at identifying weaker sources and delivering more relevant results.
The frequency of hallucinations appears to have decreased, which, despite reservations, is viewed positively as users recognize the need for careful fact-checking.
Key Takeaways
▽ Users find GPT-5 less customizable, affecting engagement.
△ Recent reports show increased hallucinations, pushing some to use Google.
▲ Multilingual support needs significant improvement, especially for Cyrillic languages.
Open discussions continue as the community pushes for a better sampling of customization options and enhanced performance in multiple languages. With many eager for adjustments, the essential question remains: Will developers respond effectively to this feedback? Only time will tell.