OpenAI eliminates ChatGPT’s typical routing scheme for most users

OpenAI quietly It reflected a major change in how hundreds of millions of people use it ChatGPT.

At a low level A blog that tracks product changesthe company said it was rolling back ChatGPT’s model router — an automated system that sends complex user questions to more advanced “inference” models — for users in the free and $5-per-month Go tiers. Instead, these users will now default to GPT-5.2 Instant, which is the fastest and most affordable version of GPT-5.2 Instant. OpenAI New model series. Free and Go users will still be able to access inference models, but will have to select them manually.

The prototype router was launched just four months ago as part of OpenAI’s push to unify the user experience with The first appearance of GPT-5. This feature analyzes user questions before choosing whether ChatGPT answers them with a responsive, cheap-to-serve AI model or a slower, more expensive logical AI model. Ideally, the router should direct users to OpenAI’s smartest AI models exactly when they need them. Previously, users accessed advanced systems through a confusing “template chooser” menu; Advantage of that CEO Sam Altman said the company hates “just as much as you do.“

In practice, the router appears to be sending more free users to OpenAI’s advanced inference models, which are more expensive for OpenAI to serve. Shortly after its launch, Altman said the router increased use of inference models among free users from less than 1 percent to 7 percent. It was an expensive bet aimed at improving ChatGPT’s answers, but the modular router was not as widely adopted as OpenAI expected.

One source familiar with the matter told WIRED that the router negatively impacted the company’s daily active users metric. While inference models are widely viewed as the frontier of AI performance, they can spend minutes working on complex questions at a much higher computational cost. Most consumers don’t want to wait, even if it means getting a better answer.

Responsive AI models still dominate general consumer chatbots, according to Chris Clark, chief operating officer at AI inference provider OpenRouter. On these platforms, he says, the speed and tone of responses tend to be paramount.

“If someone writes something down and then you have to show thinking points for 20 seconds, it’s not very engaging,” Clark says. “For general AI chatbots, you’re competing with Google (search). Google has always been focused on making search as fast as possible; it’s never said, ‘Oh, we should get a better answer, but do it slower.'”