When OpenAI launched the GPT-4o model, it became an instant favorite among users who praised the update for its accuracy and human-like tone that offered warmth and felt more conversational. So naturally, when the version was replaced with GPT-5 in ChatGPT without any warning, users felt frustrated. They immediately felt the difference not only in the performance of the update, but also in the personality. While the coding model is being presented as the most capable and advanced one yet in terms of reasoning, due to the lack of human element felt, users cannot keep themselves from missing the feeling of the older version, and hence, prompting widespread criticism and corrective measures.
When OpenAI officially launched GPT-5 on August 7, 2025, it was immediately made the default model for all logged-in ChatGPT users. The rollout had many versions tailored to different user needs including GPT-5, GPT-5 mini, and GPT-5 nano, but what should have been an update that was to be talked about for the advancement it brought forth in terms of better reasoning and more capabilities, became the talk of the town for the performance issues and the lack of personality and warmth felt in comparison to previous models like GPT-4o. OpenAI, upon getting to know about the backlash it had been receiving, took steps to address the ongoing concerns and reinstated GPT-4o as an additional option for the Plus subscribers, giving more choice to selected users.
Additionally, the GPT-5 model was , including options to select different modes such as Fast, Auto, and Thinking to better adapt to user needs and preferences and give more nuanced responses. The AI company is working ardently to bring the best elements of the previous model into the new version and retain what users loved about the model while ensuring GPT-5 offers better speed and reasoning and is a big improvement over the older versions.
Sam Altman recognized the oversight and the need for more personalization and gave users hope that the model would evolve to feel warmer without crossing boundaries, just like GPT-4o did, which ended up being annoying. The criticism received also highlights a broader issue in AI model development: technical advancement without users feeling connected to it would not serve, and it may end up alienating the users it was meant to bring utility to.
It also sheds light on another aspect of AI evolution: striking a balance between innovation and familiarity, given how rapidly upgrades are being introduced. Technical improvements should not come at the cost of user expectations and experience. OpenAI is going forward smartly by adapting and listening to the community for which it builds its technology.