ChatGPT’s model picker is back, and it’s complicated

When the opeina is turned on GPT -5 Last week, the company said that the model would make ChatGPT experience easier. Open was hoped that GPT -5 would act as “Fiting all in one size” with a router, which would automatically decide on how the user could answer the user’s question automatically. The company said OpenAI CEO Sam Altman says he hated – To pick up a version of ChatzPT that provides the right kind of response.

However, it looks like GPT -5 Unified AI Model Open Open hoped that it would happen.

Ultman says a X posts The company that introduced the new “Auto”, “fast” and “thought” settings for GPT -5 on Tuesday, which all ChatzPT users can select from the model picker. The auto setting looks like a model router of GPT -5 which Open initially announced; However, the company is also offering options to serve users, so they allow them to access directly AI models directly.

In addition to the new mode of GPT -5, Ultman said that paid users will again access several inheritance AI models GPT -4 O, GPT -4.1, and 3 co -exposed last week.

“We’re working on an update of the personality of GPT -5 that should feel warmer than the current personality but not to be boring (to most users) like GPT -4o,” wrote in the Post of Ultman X.

Chatzipt’s model picker now seems as complex as before, suggest that the GPT -5 model router is not universally satisfied user as the company expects. The expectations of the GPT -5 were high in the sky, many hoped that the Open AI model would push the GPT -4 to its limitations. However, Ruger is expected to expect the GPT -5 rollout.

GPT -4O and other AI model underestimated in ChatzPT have sparked a spark Reacts between users Who was associated with the feedback and personality of the AI models that were not expected OpenAI. In the future, Ultman has said that the company will give users a lot of advance notifications if it ever reduces GPT -4O.

The model router of the GPT -5 also seemed to have been largely broken on the day of the launch. As a result some users did not think that the AI model did not have a performance like the previous OpenAI models and forced Ultman to solve the problem in one AMA Session at Reddit. However, it looks like the router of GPT -5 may not be satisfied for all users yet.

“We’re not always going to get everything in trying #1 but I’m so proud of how fast the team can repeat,” OpenAE’s chatzipt wrote on one of Nick Tarley’s VP, Nick Tarley X posts Tuesday.

Routing requests in the right AI model are a difficult task that an AI model needs to be aligned in user preferences, as well as the specific question they are asking. The router has to decide on which the AI model should only send the prompt to a split second – that way, if a prompt goes to a quick reaction AI model, the response may still be faster.

More extensively, some people display preferences for AI models that exceed fast or slow reactions. Some users may prefer the verbosty of an AI model, while others may appreciate the opposite answers of the other.

Human attachment with some AI models is a relatively new concept that is not well understood. For example, hundreds of people in San Francisco recently Held a funeral For anthropic AI model, Claud 3.5 Sonnet, when it was taken offline. In other cases, the AI chatbats seem to be contributing to the bottom of mentally unstable people Psychological rabbit holes.

It looks like OPENY AI models have more work to do in the vicinity of alignment in separate user preferences.