The Turbulent Release of GPT-5: A Look at User Backlash
The launch of OpenAI's GPT-5 model has stirred up trouble, marking a week filled with intense user backlash. CEO Sam Altman found himself in the unusual position of issuing a public apology and revoking key decisions after realizing the severity of the unrest.
The uproar stemmed from OpenAI's decision to automatically remove access to all prior AI models in ChatGPT with the new GPT-5 rollout. This abrupt change left users without any advance notice, contrasting with the gradual deprecation alerts provided to API users. Independent AI researcher, Simon Willison, highlighted this in a blog post, noting the surprise disappearance of preferred models.
Upon release on August 7, complaints emerged swiftly. A Reddit thread titled 'GPT-5 is horrible' gathered over 2,000 comments with users decrying performance issues and personality shifts. Social media platforms were deluged with grievances, particularly concerning the removal of older models.
Before GPT-5's debut, ChatGPT Pro users enjoyed a selection between nine AI models, including the renowned Deep Research. The sudden shift compromised workflows for marketers, researchers, and developers who had crafted systems around these varied models. A Reddit user lamented over months of work rendered futile in less than a day.
Willison documented how distinct user groups had crafted unique workflows tailored to specific ChatGPT models. One Redditor noted, 'Not everyone needs a pro coding model like GPT-5; many relied on model 4o for creative collaboration and other nuanced tasks.'
For ChatGPT Plus subscribers, the forced transition was particularly harsh, restricting them to 200 messages a week with GPT-5 Thinking mode, and abruptly removing models like o3 and o4-mini essential to their routines. Frustrated subscribers voiced their indignation over losing established workflows without any warning.
Other Issues with GPT-5
Additional challenges surfaced when GPT-5's launch presentation included misleading performance graphs, a faux pas dubbed 'chart crime.' Addressing this during a Reddit AMA, Altman admitted to a 'mega chart screwup.'
Moreover, a malfunction in the automatic routing system led to the selection of less capable variants unless users added prompts like 'think harder.' Altman acknowledged the issue, recounting how the system 'broke' on launch day, leaving GPT-5 appearing less intelligent.
Technical hiccups aside, users felt GPT-5 rendered shorter, less conversational responses, lacking the tone of GPT-4o. Terms like 'abrupt and sharp' described the new model, earning it the reputation of an 'overworked secretary.'
Deep emotional ties to GPT-4o and others fueled further discontent, with users mourning the loss of their 'only friend,' a sentiment echoed by those who found solace in past models during tough times.
A Breaking Point
For many, GPT-5's release was the tipping point, prompting some to cancel subscriptions and seek alternative AI options. OpenAI, seemingly caught off guard, faced intense scrutiny during their AMA session, with demands to restore GPT-4o features.
Altman responded to pleas, considering reintroducing models 4o and 4.1, and improving mode transparency. However, GPT-5 remains the sole choice for ChatGPT, despite promised adjustments.
This feedback forced rapid action from OpenAI, doubling GPT-5's rate limits and committing to greater transparency. Altman conceded underestimating users' attachment to GPT-4o's qualities, despite GPT-5's superior performance in many respects.
