Sam Altman addresses ‘bumpy’ GPT-5 rollout, bringing 4o back, and the ‘chart crime’

During the Reddit ASK-Me -ithing session on Friday, the CEO of Openai Sam Altman and the main members of the GPT-5 team were heated with questions about the new model and asked for its previous model, GPT-4O.

They also asked Altman about the most embarrassing – and perhaps the funniest – Snafu in the presentation, “The Graphic Crime”.

One of the new features offered by GPT-5 is a actual time router that decides the model that must be used for a specific demand, either responding quickly or spending additional time “thinking” through answers.

But many people in Ama on R/Chatgpt Reddit complained of GPT-5 also did not work for them as 4O did. Al-Taman said the reason that the GPT-5 looked “Dumber” was that the router was not working properly when it was launched on Thursday.

“GPT-5 will look more intelligent starting today. Yesterday, we had SEV, and Autoswitcher was outside the committee for a piece of today, and the result was the GPT-5 seemed way.

However, the people on Ama pressed hard to return 4O to the subscribers in addition to that altman promised them at least in consideration. “We are looking to allow users in addition to continuing to use 4O. We are trying to collect more data about the bites,” Books.

Altman also promised, “We’ll write down modified limits for excess users while finishing starting.” These people should give an opportunity to play and learn the new model, and adopt us use cases without worrying about running out of monthly claims.

As expected, he was also asked about the inaccurate chart that the team made during the direct presentation that soon became many jokes “plan crimes”. Give the graph less than the standard with a much longer tape.

Altman did not answer the questions about the graph during Ama, but Thursday called the “Screwup Mega Chart” chart on X. Others noticed that the charts in the blog post were correct.

But the damage was done. This was followed by jokes about the use of GPT for plans in the corporate view. The GPT-5 references, Simon Wilison, who had early access and generally admired the performance of the form, indicated that converting data into a schedule was “a good example of GPT-5”.

However, Altman promised repairs to the elements that seemed to be more interested in. Ama ended with a promise: “We will continue to work to get stable things and we will continue to listen to the comments.”