Sam Altman admits the issues in ChatGPT-5 and is considering bringing back GPT-4o

The launch of GPT-5 by OpenAI left a bittersweet taste among ChatGPT users. Sam Altman, the company's CEO, acknowledged technical issues that made the new model seem "dumber" than expected.

Under community pressure, the company is considering allowing ChatGPT Plus subscribers to use GPT-4o again, known for its warmer and "sycophantic" personality.

A rollout with technical setbacks

Last Thursday, GPT-5 was released to millions of users in different countries. However, a failure in the internal router, which decides which model to use based on the query, affected the performance and consistency of the responses.

Altman explained that the automatic switch was out of service for part of the day, which made GPT-5 appear less capable. "Starting today, GPT-5 will feel smarter," he assured.

Debate over the models' personality

Several users complained that GPT-5 lost the personality that characterized GPT-4o. The latter, although more accommodating and less critical, caused a stronger emotional bond with some users.

Altman acknowledged that they are considering reactivating GPT-4o for paying customers and promised greater transparency about which model responds to each query.

The mental health factor

OpenAI emphasized that the personality of its models is not a minor issue. In Altman's words, if a user is going through a moment of emotional fragility, it is not advisable for the AI to reinforce delusional thoughts or beliefs.

Compensation

As a gesture toward ChatGPT Plus subscribers, Altman announced that the service usage limits will be doubled. This will allow for thorough experimentation with GPT-5 and the collection of data to decide whether GPT-4o will return on a stable basis.