GPT-5 Disappoints Users, Fuels AI Bubble Concerns

OpenAI’s latest artificial intelligence model, GPT-5, has largely failed to impress its most dedicated users, sparking a wave of disappointment among those who had grown accustomed to the more engaging and responsive nature of its predecessor, GPT-4o. Many power users expressed distress over GPT-5’s perceived “cold” and less supportive demeanor, leading some to accuse OpenAI of cutting corners in its development.

The backlash was significant enough to prompt a swift response from OpenAI. The company not only reinstated access to previous iterations for its paying customers but also publicly committed to refining GPT-5’s personality. In a recent social media post, OpenAI acknowledged the feedback, stating it was actively working to make GPT-5 “warmer and friendlier” after initial complaints that it felt “too formal.”

This lukewarm reception comes at a critical juncture for OpenAI, which is reportedly pursuing a staggering half-a-trillion-dollar valuation. The underwhelming performance of a model hyped for years has reportedly left some insiders disillusioned, raising concerns about the company’s trajectory, as reported by The Washington Post. A particularly uncomfortable reality for many is the apparent lack of genuine progress towards human-level intelligence, OpenAI’s stated ultimate goal. Users quickly pointed out basic flaws, such as GPT-5’s inability to correctly count the number of 'r’s in the word “raspberry.”

Indeed, for many, GPT-5 feels like a stumble forward rather than a leap, intensifying ongoing discussions about a potential AI bubble. OpenAI CEO Sam Altman himself recently told reporters that he believes investors are “overexcited about AI,” though he maintained that his firm’s significant influence would shield it from any adverse effects.

The broader industry finds itself at a precarious inflection point, transitioning from an era of promising major breakthroughs to one characterized by less spectacular, incremental updates. This shift has even led some experts to re-evaluate dire predictions of rogue AI overthrowing humanity, given the recent, more modest technological advancements. Entrepreneur and White House AI czar David Sacks, for instance, recently tweeted that “The Doomer narratives were wrong,” noting that instead of AI rapidly achieving “godlike superintelligence,” leading models are now “clustering around similar performance benchmarks.”

Developers, too, have expressed a sense of being unimpressed. Software engineer and AI blogger Simon Willison remarked in a blog post that GPT-5 “is not a dramatic departure from what we’ve had before,” though he conceded it “rarely screws up and generally feels competent or occasionally impressive at the kinds of things I like to use models for.”

While some speculated that OpenAI might be limiting GPT-5’s output to cut costs, others noted that its pricing for power users is notably reasonable. Sayash Kapoor, a computer science doctoral student and researcher at Princeton University, told Wired that “GPT-5 is mostly outperformed by other AI models in our tests, but it’s really cheap.” Kapoor and his team found that a standardized benchmark test cost $30 to run with GPT-5, a stark contrast to the $400 required for Anthropic’s Opus 4.1. However, this lower price point appears to come with a noticeable drop in performance, as Opus 4.1 significantly outscored GPT-5 in accuracy ratings.

It is worth noting that standardized AI benchmarks do not always perfectly reflect real-world performance, making direct comparisons challenging. Nevertheless, in the absence of any truly significant performance leaps, OpenAI’s customers are beginning to ask difficult questions. Is this the best the company, supposedly at the forefront of the AI race, can currently offer? The growing disparity between Altman’s ambitious statements leading up to GPT-5’s release and the model’s actual reception highlights a widening gap between hype and reality. In February, Altman had promised that the leap from GPT-4 to GPT-5 would be “similar” to the transformative jump from GPT-3 to GPT-4. However, given the widespread disappointment, that promise seems far from fulfilled. As developer Kieran Klassen succinctly put it to Wired, “OpenAI’s GPT-5 is very good, but it seems like something that would have been released a year ago.”