OpenAI’s GPT‑5 landed last week to predictable hype. Headlines focused on how much faster and sharper it is, the shiny benchmarks, and rollout to hundreds of millions of ChatGPT users.
But outside the clickbait headlines, the tone is different. Yes, the launch is big, but for many people in the field, it’s not the leap the OpenAI suggests.
User backlash over model removal
OpenAI’s decision to automatically withdraw all previous ChatGPT models upon GPT-5’s release blindsided many users. This removed favoured AI personalities and disrupted workflows without prior notice, sparking one of the most intense user backlashes in ChatGPT’s history.
Many expressed emotional attachment to older versions such as GPT-4o, describing them as more empathetic and reliable companions. Following this uproar, OpenAI quickly reinstated GPT-4o for Plus subscribers and added an option to toggle legacy models in settings, signalling recognition that a single “best” model does not meet all needs.
Performance issues and technical glitches
Despite GPT-5’s praised improvements, early users reported slower response times, unexpected basic errors, and a perceived loss of creativity and nuance compared to GPT-4o.
A critical bug in the model routing system led GPT-5 to behave as a less capable model, occasionally undermining confidence. CEO Sam Altman publicly acknowledged these issues and committed to updates, including increased rate limits and enhanced “thinking modes” to improve accuracy and reliability.
Changes in safety and tone
GPT-5 reduces overly flattering responses, adopting a more measured and business-like tone. While this improves factual reliability and reduces bias, some users found the interactions colder and less engaging.
The model also introduced “safe completions” to handle sensitive queries, offering constrained but helpful answers rather than outright refusals, enhancing user experience while managing risks.
Coding and creative limitations
In coding tasks, GPT-5 showed mixed results, with some users reporting broken scripts, inconsistent outputs, and weaker performance than prior versions.
Despite stronger analytical coding features highlighted by OpenAI, many developers faced frustrations that impacted workflows, sparking calls for clearer communication and fixes from the company.
