When OpenAI launched GPT-5 last week, it positioned the model as a “one size fits all” solution. The company said this unified approach would eliminate the need for users to navigate its model picker — a long, complicated menu of AI options that OpenAI CEO Sam Altman has publicly said he dislikes. The AI was designed with a real-time router to automatically decide whether a query should be answered quickly or processed more thoroughly—eliminating the need for a long model picker menu.

However, on launch day, that router was not functioning properly, leading some users on Reddit to describe GPT-5 as performing worse than its predecessor GPT-4o.

In a Reddit AMA on Friday, CEO Sam Altman explained:
"GPT-5 will seem smarter starting today. Yesterday, we had a sev and the autoswitcher was out of commission for a chunk of the day, and the result was GPT-5 seemed way dumber." He added that OpenAI is adjusting “how the decision boundary works” so users get the right model more often and will make it more transparent which model is answering.

Just days after launch, Altman announced on X that ChatGPT users now have Auto, Fast, and Thinking modes in the model picker—alongside optional access to legacy AI models.

For paid subscribers, GPT-4o, GPT-4.1, and o3 are once again available. GPT-4o appears by default in the picker, while other models can be added in settings.

Altman acknowledged user frustration at the sudden removal of GPT-4o, saying the company will provide “plenty of advance notice” if it ever deprecates the model in the future.

In the AMA, users lobbied for GPT-4o’s return, arguing it better fit their needs. Altman responded: "We are looking into letting Plus users continue to use 4o. We are trying to gather more data on the tradeoffs."

He also promised to double rate limits for Plus users to give them more opportunity to test GPT-5 without hitting usage caps.

Alongside functionality changes, Altman said the company is working on a warmer personality for GPT-5—"We are working on an update to GPT-5’s personality which should feel warmer than the current personality but not as annoying (to most users) as GPT-4o,” Altman wrote in the post on X. “However, one learning for us from the past few days is we really just need to get to a world with more per-user customization of model personality.”

The GPT-5 launch event also drew attention for an embarrassing visual error dubbed the “chart crime.” A presentation slide showed a much taller bar for a lower benchmark score, quickly becoming a social media joke.

Altman did not address the issue in the AMA, but had called it a “mega chart screwup” on X the previous day. While the published blog post contained correct charts, the incident fueled lighthearted criticism and became part of early GPT-5 chatter.

OpenAI executives acknowledge that routing prompts effectively remains complex. Nick Turley, OpenAI’s VP of ChatGPT, noted on X: “We’re not always going to get everything on try #1 but I am very proud of how quickly the team can iterate.”

Routing decisions must balance user preferences, prompt type, and model capabilities—often in milliseconds. But user attachment to specific models can go beyond speed or accuracy, with some preferring a certain style or personality.

Such attachment has become a cultural phenomenon: TechCrunch noted that hundreds of people in San Francisco even held a funeral for Anthropic’s Claude 3.5 Sonnet after it was taken offline.

Altman ended the AMA saying, “We will continue to work to get things stable and will keep listening to feedback.”

Q: Why did GPT-5 seem “dumber” at launch?
A: A malfunctioning model router caused performance issues on launch day.

Q: What changes has OpenAI made since?
A: Restored GPT-4o and other legacy models, reintroduced the model picker, and added Auto, Fast, and Thinking modes.

Q: Will Plus users get more access?
A: Yes, rate limits for Plus subscribers will be doubled during rollout.

Q: Is GPT-5’s personality changing?
A: Yes, OpenAI is working on a warmer personality and exploring per-user customization.

Q: What was the “chart crime”?
A: A visual error in the launch presentation that showed an incorrect bar chart, later corrected in the blog post.

The GPT-5 rollout reflects both the promise and complexity of unifying AI capabilities under a single model. While the router system aims to simplify user experience, early glitches and strong attachment to previous models like GPT-4o show that technical efficiency isn’t the only factor driving user satisfaction.

By restoring legacy models and adding personality customization, OpenAI is signaling it values user trust and flexibility alongside technological advancement. The early turbulence may slow GPT-5’s push toward a streamlined AI experience, but it also offers lessons in how to balance innovation with familiarity in a rapidly evolving AI ecosystem.

Editor’s Note: This article was created by Alicia Shapiro, CMO of AiNews.com, with writing, image, and idea-generation support from ChatGPT, an AI assistant. However, the final perspective and editorial choices are solely Alicia Shapiro’s. Special thanks to ChatGPT for assistance with research and editorial support in crafting this article.