GPT-5's Launch Marred by User Frustration, OpenAI Promises Fixes

OpenAI's latest model, GPT-5, has encountered unexpected initial feedback from early users who describe its performance as "clumsy," even falling short of its predecessor. Many users have reported that the model, which was touted to elevate ChatGPT to a "doctoral expert" level, frequently makes errors in simple mathematical and spelling tasks and sometimes fabricates information. This has led to widespread disappointment among paying users, with some calling for a return to the previous model, GPT-4o.

The root of the negative feedback lies in a new mechanism introduced in GPT-5: an "automatic switcher" designed to optimize computational resources by selecting different models based on the complexity of user queries. However, this mechanism has caused confusion and dissatisfaction. For instance, when asked how many letters "b" are in the word "blueberry," GPT-5 initially gave the incorrect answer "three." Only after being prompted to "think again" did it provide the correct response. This lack of transparency has been a significant source of user frustration.

In response to the backlash, OpenAI's management has acknowledged the technical issues and promised improvements. The company's CEO quickly addressed the concerns, attributing the problems to a malfunctioning "automatic switcher" that led to the system using weaker models. This technical glitch resulted in a subpar user experience, making GPT-5 appear "much clumsier." The CEO also announced plans to allow Plus paying users to continue using GPT-4o and to double the usage rate limit for Plus users to encourage better adaptation to the new model.

Despite the negative feedback, some professional users have given GPT-5 positive reviews. One developer praised it as a model that "can handle the job" and occasionally performs exceptionally well, though it lacks fundamental differences from previous models. Another expert was impressed by its capabilities in research, writing, and simplifying programming tasks. However, these positive reviews are overshadowed by the widespread criticism and technical issues reported by early users.

The release of GPT-5 comes at a critical time for OpenAI, as the company faces intense competition in the AI field. Maintaining its leadership position and convincing users to pay for its advanced services is crucial for OpenAI to offset its significant investments in talent, chips, and data centers. The unstable launch of GPT-5 has cast a shadow over these commercial goals, raising concerns about the model's reliability and performance.

OpenAI's efforts to address the issues include acknowledging a "major chart error" during the launch presentation, which further highlighted the chaotic and rushed nature of the release. The company is now focusing on improving the model's performance and user experience, with the CEO promising that "from today on, GPT-5 will appear smarter." The company is also exploring ways to allow Plus users to continue using GPT-4o, providing them with double the usage rate limit to encourage better adaptation to the new model.

In summary, while GPT-5 represents a significant advancement in AI technology, its initial user feedback has been overwhelmingly negative. The model's clumsy performance and inaccurate information have led many users to demand a return to the previous version, GPT-4o. OpenAI's decision to bring GPT-4o back online for Plus and Team users is seen as a temporary solution to address these concerns. The company will need to continue refining and improving GPT-5 based on user feedback and technical evaluations to meet the high expectations set by its previous models.