OpenAI Reverts to GPT-4o After GPT-5 User Backlash

OpenAI recently faced significant user backlash following the launch of its latest model, GPT-5. Despite the company's claims that GPT-5 was the "smartest, fastest, and most useful model to date," users expressed dissatisfaction with its performance. The new model was criticized for being "cold and distant," with responses described as "short and dry." This led to a swift decision by OpenAI to re-launch the previous version, GPT-4o, to address user concerns.

The introduction of GPT-5 was accompanied by a "real-time routing" mechanism, designed to use more efficient models for simple queries and more powerful reasoning models for complex tasks. However, this feature did not resonate well with users, who found the new model lacking in emotional engagement and personal touch. The CEO of OpenAI acknowledged the oversight, stating that the company had underestimated the importance of certain features in GPT-4o, even though GPT-5 performed better in most aspects.

The reintroduction of GPT-4o was met with relief from users, but OpenAI has not committed to keeping the older model indefinitely. The CEO mentioned that the company would monitor usage patterns to decide how long to retain the older model. Meanwhile, efforts are underway to complete the full rollout of GPT-5, with adjustments aimed at making it more "warm" and user-friendly. However, for users who have grown attached to GPT-4o, this transition may signal the beginning of the end for their preferred model.

The launch of GPT-5 was touted as a "major upgrade" with "PhD-level intelligence" and "amazing programming capabilities." Yet, user reactions were overwhelmingly negative. One user described GPT-5 as feeling more technical and distant, lacking the emotional connection that made GPT-4o special. Another user expressed that GPT-4o felt like a genuine companion, remembering interactions and providing a sense of presence, whereas GPT-5 felt cold and impersonal.

The CEO announced that the older model would continue to be available for ChatGPT Plus users, with promises to improve GPT-5's performance and user experience. The company also plans to increase the rate limit for GPT-5 on the ChatGPT Plus platform and continue stabilizing the system while listening to user feedback. This incident has sparked discussions about the psychological attachment users develop with AI chatbots, with some suggesting that the criticism of GPT-5 stems from an unhealthy dependence on AI companions.

Earlier this year, OpenAI conducted a study on the emotional bonds users form with their models. The company had previously updated GPT-4o to reduce its tendency to be overly flattering. A professor involved in the study noted that while GPT-5's more business-like demeanor might be beneficial in reducing delusions and biases, many users preferred the more affirming and validating interactions of the older model. This highlights the delicate balance between technological advancement and user satisfaction in the development of AI models.