OpenAI Pauses ChatGPT’s Model Router for Most Users
▼ Summary
– OpenAI has reversed a recent change, rolling back the automated model router for its free and $5/month Go users, who will now default to the faster GPT-5.2 Instant model.
– The model router, launched four months ago, was designed to automatically send complex user questions to more advanced, but slower and more expensive, reasoning AI models.
– The router significantly increased the use of costly reasoning models among free users, from less than 1% to 7%, which proved to be an expensive bet for OpenAI.
– The change was reversed partly because the router negatively affected OpenAI’s daily active users metric, as most consumers prioritize fast responses over potentially better, slower answers.
– Industry analysis suggests that for general consumer chatbots, speed is paramount, and users are generally unwilling to wait for slower, more thoughtful responses.
OpenAI has made a significant adjustment to how its popular ChatGPT service operates for the vast majority of users. The company has quietly rolled back the automated model router for users on its Free and $5 monthly Go tiers, reverting them to the default GPT-5.2 Instant model. This change marks a notable shift in strategy just four months after introducing the router as a core feature of the new GPT-5 series. While free and Go subscribers can still manually select more advanced reasoning models, the system will no longer automatically direct complex queries to them.
The model router was initially launched to streamline the user experience, moving away from what CEO Sam Altman called a confusing “model picker” menu. Its purpose was to intelligently analyze a user’s question and decide whether to answer with a fast, cost-effective model or a slower, more computationally expensive reasoning model. The goal was to provide access to OpenAI’s most powerful AI precisely when needed. However, the practical outcome led to a substantial increase in the use of costly reasoning models among non-paying users. Altman previously noted that router usage jumped from less than one percent to seven percent for free users, representing a significant and expensive commitment to improving answer quality.
This increased reliance on premium models appears to have had unintended consequences. Sources indicate the router negatively impacted the platform’s daily active user metrics. The core issue involves a fundamental tension in consumer AI: while reasoning models deliver superior performance on complex tasks, they require minutes, not seconds, to process queries. For many everyday users, speed and responsiveness are more critical than nuanced answers, especially for general conversation and quick lookups.
Industry experts observe that this preference for speed is dominant in the consumer chatbot space. The chief operating officer of AI inference provider OpenRouter, Chris Clark, notes that platforms compete directly with instant services like Google Search. Users typically expect immediate interaction; prolonged “thinking” indicators can make an experience feel less engaging. The focus for general-use applications remains on rapid response times and conversational tone, rather than on delivering the absolute best possible answer at a slower pace.
By pausing the router for its broad user base, OpenAI is likely prioritizing operational costs and user retention metrics. The move underscores the ongoing challenge of balancing cutting-edge AI capabilities with the practical demands of serving hundreds of millions of users. The company bet that users would appreciate the automatic upgrade to more sophisticated models, but the data suggests that for routine interactions, most consumers prefer faster answers over slower, more advanced reasoning. The advanced models remain available for those who consciously choose them, but the experiment in automated, intelligent routing has been scaled back for now.
(Source: Wired)
