Sam Altman Altstman ‘Bumpy’ Gpt-5 Rollout, making 4o Back, and ‘Cruis Cruis’

During Reddit Ask me-nothing Sessions on Friday, COo Opening Operanai Sam Altman and the Key of the GPT-5 team opens with questions about the new model and the request to bring the previous model, GPT-4O.

He also asked Altman about the most shy – and perhaps the funniest – Snafu at the presentation, “said” criminals. “

One of the new features rolled in GPT-5 are Real Time Router That resolves the models used for certain directions, react quickly or take additional time to “think” through the answers.

But many people in ama in R / Chatgppt Reddit complain Gpt-5 doesn’t work well for them as 4o. Altman says that the reason is looking for “a dumber” is a route not working properly when you roll on Thursday.

“GPT-5 will look fresh now. Yesterday, we have Sev and the autoswitcher could not be used to make a precise model. We will make the model more. We will make the model more.

Still, the people of the Ama in the friday has hard to carry 4o back for customers plus that they promised with the most unknown altman. “We are looking for users to use 4o. We try to collect more data on the trade,” he said.

And Altman also promises also promise, “We are going to the second level limit for users when we complete the rollout.” This should give you the opportunity to play and learn new models, with case of useless use without worrying.

I hope, he also asked for a very accurate chart of the team during a quick presentation of the assassination of a large number of ass “Chinese jokes “jokesSee rankings-. Bagan displays lower benchmark score with a higher bar.

Altman did not answer the question about the chart when the Father, but Thursday was calling her chart as “the thread Chart Strewup” ing xSee rankings-. Other note the ladder in Post blog administered is trueSee rankings-.

But the damage is done. The joke begins using GPT for charts in the company presentation. GTPT-5 Reviewer Simon Million, who has an early access and is generally pleased with the model performance, also pointed The data converts the data to the table is “good examples to fail GPT-5 failure.”

In any case, Altman promises fix items that seem to concern others. They ended a fasm with the promise of “we will continue to work to get stabilize and they will keep listening.”