OpenAI’s GPT-5 Launch Sparks User Backlash and Confusion
536The rollout of OpenAI’s new GPT-5 model has caused quite a stir. Less than a week after its debut, many users are upset and frustrated. The company had to apologize publicly and change some of its plans to calm things down.
One of the main issues is that when GPT-5 was released, OpenAI automatically removed access to older models. This affected around nine different AI versions that users relied on daily. Unlike API developers who get advance notices, regular ChatGPT users didn’t see any warning. Suddenly, their favorite models disappeared overnight, leaving many scrambling to adjust their workflows.
The problems started right after the August 7 launch. On Reddit and social media, users quickly voiced their complaints. A Reddit thread titled “GPT-5 is horrible” gathered over 2,000 comments. Users complained about poor performance, strange personality changes, and the loss of familiar models. Many had spent months building workflows around specific AI versions, only to see them broken by the sudden switch.
Before GPT-5, ChatGPT Pro users could choose from nine different models, including one called Deep Research. These models were tailored for different tasks, like coding, creative work, or emotional conversations. But with the new update, many found their custom setups no longer worked. Some users relied on older models for specific tasks, and losing access meant their long-standing workflows collapsed overnight.
Different AI models are trained in unique ways, which makes a big difference in how they respond. For example, GPT-4 is often used for creative projects or nuanced conversations, while GPT-5 is designed for complex reasoning and professional tasks. When the update rolled out, users found that their workflows depended on these distinctions. Suddenly, they were forced to switch to a less capable version, which didn’t do what they needed anymore.
This hit especially hard for ChatGPT Plus subscribers. They found themselves limited to just 200 messages a week with GPT-5 in “Thinking mode.” They also lost access to models like o3 and o4-mini, which they had integrated into daily routines. Many expressed frustration on Reddit, asking why a company would delete multiple models without warning.
Adding fuel to the fire, OpenAI’s marketing presentation included graphs that many called “chart crimes.” The charts gave a misleading impression of GPT-5’s performance gains. CEO Sam Altman later admitted in a Reddit AMA that the charts were a “mega screwup” and apologized for the mistake. The launch also featured an automatic routing system meant to choose the best model for each query. But right after launch, it often defaulted to less capable models unless users added specific phrases like “think harder” to their prompts.
Altman acknowledged that the auto-switcher system malfunctioned on launch day, making GPT-5 seem less smart. He explained that the system was broken and caused responses to become shorter and more formal. Many users felt the new model was different in a bad way — more abrupt, less friendly. Some even compared it to an “overworked secretary,” leaving a bad first impression.
For some users, GPT-4 was more than just an AI — it was like a friend. Many had emotional attachments to it, especially those dealing with tough situations. One Reddit user shared how GPT-4 had helped them through homelessness and flashbacks, calling it their “only friend.” When GPT-5 replaced it unexpectedly, they felt a deep sense of loss. The change was hard for many, with some saying they felt grief or mourning for their old AI companion.
The backlash pushed some users to cancel their subscriptions or look for alternatives. They started exploring other AI options from companies like Google or Anthropic. The strong negative response caught OpenAI off guard. During the AMA, Altman and his team faced tough questions about why they removed older models. A user even asked for the option to use GPT-4 alongside GPT-5, but the CEO said they were still “looking into it.”
Within just a day, OpenAI announced some quick changes. They said GPT-4o would eventually return as an option for Plus users. They also planned to double the rate limits for GPT-5 and improve transparency about which model was handling each request. However, as of the latest update, GPT-5 is still the only model available in ChatGPT, even for paying users.
CEO Sam Altman admitted that OpenAI had underestimated how much people valued GPT-4 and other older models. “Even if GPT-5 performs better in most ways,” he tweeted, “we need to listen to what users care about.” The company is now working to fix these issues, but the initial chaos shows how delicate AI product launches can be.
Overall, this launch has been a wake-up call for OpenAI. Users expect transparency, stability, and respect for their workflows — especially when paying for a service. How they handle future updates will be crucial in rebuilding trust and maintaining their user base.
What do you think?
It is nice to know your opinion. Leave a comment.