OpenAI released its GPT-5 model five days ago, prompting immediate scrutiny from tech media and users regarding its performance compared to prior expectations and previous iterations.
Initial assessments of GPT-5 have been critical. Axios described the new large language model as having “landed with a thud,” while Ars Technica characterized the launch as “messy.” The Information suggested that OpenAI may have entered the “trough of disillusionment,” a phase where initial high expectations for a technology are replaced by disappointment as real-world results do not align with the hype. This sentiment indicates that the model’s performance has not met the elevated promises made by OpenAI.
Beyond media reviews, social media users have also criticized GPT-5, with some suggesting the model represents a significant regression. Speculation online included theories that OpenAI might be attempting to reduce operational costs by limiting the model’s capabilities. Developers also reported functional issues; multiple developers informed The Information that GPT-5 exhibited difficulties in autonomously determining when to “think harder” when processing more complex prompts. This particular issue has been observed by “power users” of the platform, who often engage with the AI using advanced queries.
The perceived improvements in GPT-5 have been described as incremental, especially when compared to other recent model releases from OpenAI and its competitors since the launch of GPT-4o in May. This incremental nature of the improvements contributes to the overall user dissatisfaction. Users have reported that GPT-5 provides responses that are shorter and less precise compared to previous versions. The inability for all users to revert to a previous model has intensified this criticism. Only paid subscribers of OpenAI’s services retain the option to switch back to GPT-4o, leaving other users without recourse and potentially affecting user retention.
