GPT-5 disappoints users with “cold” responses as OpenAI restores older models

OpenAI’s GPT-5 has disappointed power users and developers who found the model to be “cold,” less capable than expected, and failing to deliver the dramatic improvements CEO Sam Altman had promised. The lukewarm reception has forced OpenAI to backtrack on design choices and restore access to previous model versions, raising questions about whether the company can justify its projected half-trillion-dollar valuation amid growing concerns about an AI bubble.

What you should know: GPT-5’s release has been marked by widespread user dissatisfaction and performance concerns that fall short of OpenAI’s ambitious promises.

• Users complained about the model’s “cold” and formal demeanor compared to GPT-4’s more supportive tone, prompting OpenAI to tweet that it’s “making GPT-5 warmer and friendlier based on feedback that it felt too formal before.”
• The model fails basic tasks like correctly counting the number of r’s in “raspberry,” highlighting gaps in fundamental reasoning capabilities.
• OpenAI was forced to restore access to previous model iterations for paying customers due to user pushback.

The big picture: GPT-5’s underwhelming performance reflects broader industry challenges as AI development shifts from revolutionary leaps to incremental improvements.

• Sam Altman, OpenAI’s CEO, acknowledged that investors are “overexcited about AI,” though he claimed OpenAI remains insulated due to its market influence.
• The disappointing release comes after years of hype, with Altman previously promising the GPT-4 to GPT-5 leap would be “similar” to the transformative jump from GPT-3 to GPT-4.

Performance benchmarks: While GPT-5 offers competitive pricing, it trails behind competitors in accuracy and capability tests.

• Princeton researcher Sayash Kapoor found that standardized benchmark tests cost $30 with GPT-5 versus $400 with Anthropic’s Claude Opus, but noted that “GPT-5 is mostly outperformed by other AI models in our tests.”
• Anthropic’s Claude Opus significantly outperformed GPT-5 in accuracy ratings, highlighting the trade-off between cost and performance.

What they’re saying: Industry experts and developers express measured disappointment with GPT-5’s capabilities relative to expectations.

• “It’s not a dramatic departure from what we’ve had before — but it rarely screws up and generally feels competent or occasionally impressive,” wrote AI blogger Simon Willison.
• Developer Kieran Klassen told Wired: “OpenAI’s GPT-5 is very good, but it seems like something that would have been released a year ago.”
• David Sacks, an entrepreneur and White House AI advisor, tweeted: “The Doomer narratives were wrong,” noting that instead of achieving “godlike superintelligence,” leading models are “clustering around similar performance benchmarks.”

Why this matters: GPT-5’s reception suggests the AI industry may be hitting practical limitations that could deflate inflated expectations and valuations.

• The gap between OpenAI’s marketing promises and actual delivered capabilities raises questions about the sustainability of current AI investment levels.
• Expert predictions about AI risks are being recalibrated as models show more incremental rather than exponential improvements toward artificial general intelligence.