‘We totally screwed up’: Sam Altman explains what went wrong with GPT-5 launch

OpenAI CEO Sam Altman has acknowledged the mistakes made by the AI startup during its latest GPT-5 model rollout. Notably, soon after the rollout, OpenAI had discontinued all its previous models, leading to a barrage of criticism from existing users and threats of subscription cancellations.

The ChatGPT maker soon realized its mistake and began providing GPT-4o — the model that previously powered the chatbot — to ChatGPT Plus users, while also offering additional rate limits for GPT-5 Standard and Thinking models.

OpenAI had claimed that its new model was leaps and bounds ahead of previous offerings in areas like coding, reasoning, accuracy, health, and multimodal abilities. However, users complained that the new model gave shorter answers and showed less emotional depth in its responses compared to earlier models.

In the immediate aftermath of the model launch, OpenAI promised to make the new model “warmer” while also implementing a host of other changes.

Sam Altman on GPT-5’s patchy rollout:

Altman, who had been teasing the GPT-5 rollout as the next big leap in AI for months, if not years, appeared to have been surprised by the response the model received from users.

In a recent discussion with reporters on Thursday (as quoted by Bloomberg), the OpenAI chief executive admitted, “I think we totally screwed up some things on the rollout.”

“We’ve learned a lesson about what it means to upgrade a product for hundreds of millions of people in one day, and the differences in the kinds of attachment people have with this product versus previous products,” Altman added.

Altman went on to state that ChatGPT API traffic had doubled in the 48 hours following the new model launch and that ChatGPT app usage was at a “complete high” in the days after the GPT-5 release. He reportedly also agreed that it was the wrong decision to deprecate all the older models.

Last week, OpenAI announced that it had reached 700 million weekly users — a figure the company would have hoped to grow further with the launch of its new models. During its previous major GPT model launch earlier in the year, OpenAI had introduced native image generation capabilities to GPT-4o, which sparked a Studio Ghibli–style craze worldwide and led to a major spike in ChatGPT usage.

That same craze was nowhere to be seen this time around, as social media quickly filled with criticism of the approach taken with the GPT-5 launch. The company even had to restore the model picker after user backlash — notably, removing the model picker entirely had been one of Altman’s key promises for GPT-5 in the months leading up to the launch.