To kick off the live demonstration, OpenAI released its newest language model, GPT-5, and the world tuned in. The event took a sudden turn when the most senior team came up with an incorrect chart. This gaffe hence immediately inspired a flurry of jocular references to “chart crime.” In recognition of the error, OpenAI’s CEO Sam Altman referred to it as a “mega chart screwup” on social media platform X.
While presenting, Altman responded to questions about the issues users had raised and committed to fixing the most immediate problems first. He said that the company needs to be more clear about which model is answering user questions. This commitment comes in response to a recent severity event (sev) 1 related to the malfunction of GPT-5’s autoswitcher. Consequently, we ran into severely reduced performance for much of the day. Altman echoed this sentiment in explaining why under these circumstances, GPT-5 seemed “way dumber.”
Simon Willison, one of the reviewers who got early access to GPT-5, was generally impressed with the model’s performance. He noted a specific shortcoming: the handling of data tables. Willison gave this as a particularly instructive “high mark of a GPT-5 fail,” a window into the places where even artificial intelligence’s best efforts still fall short.
The new model features a built-in real-time router. The prompt you enter determines which version of GPT will be used versus others, but it does so instantly. This cutting edge also makes it so GPT-5 can choose to answer quickly or to spend more time offering nuanced and detailed responses. Altman heralded the significance of this ability, adding that interventions would be inserted to enhance decision-making in the model.
“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. Also, we are making some interventions to how the decision boundary works that should help you get the right model more often. We will make it more transparent about which model is answering a given query,” – Sam Altman
Altman’s being pretty straightforward about performance concerns. He announced that rate limits would be doubled for Plus users once GPT-5 rolls out. This shift allows participants to freely dive into the unknown aspects of the new model. Say goodbye to the stress of going over monthly prompt caps!
“We are going to double rate limits for Plus users as we finish rollout. This should give people a chance to play and learn the new model, adopt it to their use cases without worry of running out of monthly prompts,” – Sam Altman
OpenAI has been working on allowing Plus users to continue using its older, more advanced, model, 4o. Their goal is to take what they learn to better understand user preferences and trade-offs.
“We are looking into letting Plus users to continue to use 4o. We are trying to gather more data on the tradeoffs,” – Sam Altman
As OpenAI navigates the rollout of GPT-5 and addresses user feedback, Altman reassured stakeholders of the company’s commitment to stability and improvement.
“We will continue to work to get things stable and will keep listening to feedback.” – Sam Altman
