OpenAI has rolled back a significant change in how ChatGPT operates for its Free and $5-a-month Go users. The company removed the model router, an automated system that directed complex queries to more advanced AI models, and instead, these users will now default to the faster and cheaper GPT-5.2 Instant model. While free users can still access advanced reasoning models, they need to select them manually. This change comes after user feedback indicated a preference for a simpler experience, as the router had initially increased the use of reasoning models among free users from under 1% to 7%, which proved costly for OpenAI. The decision reflects ongoing challenges for OpenAI, particularly as competition with Google ramps up. Despite ChatGPT's massive user base of over 800 million weekly active users, there has been a slight decline in user engagement since new content restrictions were implemented. OpenAI's CEO, Sam Altman, acknowledged the need to improve the core consumer product in light of this competition. Data from Similarweb shows that ChatGPT's growth has stalled, with users spending less time on the platform compared to Google's Gemini. OpenAI's model router faced backlash from users who preferred a more hands-on approach in selecting models. After launching the router in August, OpenAI quickly reinstated the model picker due to user dissatisfaction. The company aims to balance increased usage with safety, especially as reports indicate that many users exhibit signs of mental health emergencies. OpenAI now claims that the GPT-5.2 Instant model performs better on safety benchmarks, leading to the decision to stop routing sensitive conversations to reasoning models. The model router remains available for paid subscribers, suggesting that OpenAI still sees value in the concept and may refine it for broader use in the future.