OpenAI has quietly reversed a core product strategy for its mainstream users, discontinuing the automated "model router" system designed to improve answer precision in favor of faster response times to retain users. This move highlights the difficult balance between advanced "reasoning" capabilities and consumer demand for instant responses in the highly competitive AI chatbot market.
According to reports, OpenAI has removed the "model router" feature from its free tier and $5/month subscription plan, with users now defaulting to the faster, lower-cost GPT-5.2 Instant model. Previously, the system automatically analyzed user queries and routed complex questions to more powerful but slower "reasoning" models. Now, regular users must manually switch to access these advanced features.
Insiders reveal this reversal stems primarily from its impact on business metrics: the router's latency negatively affected daily active users. While reasoning models represent the cutting edge of AI performance, processing complex queries can take minutes—leading to churn among mainstream users unwilling to wait.
Meanwhile, facing pressure from Alphabet's surging Gemini, OpenAI is grappling with stagnant user growth, forcing a strategic pivot toward prioritizing interaction efficiency in its core market.
An OpenAI spokesperson confirmed the change, stating that free and Go users overwhelmingly prefer the default seamless chat experience. Though the company plans to reintroduce the feature after technical refinements—and paid subscribers retain access—this rollback underscores the challenge tech giants face in integrating high-cost, high-latency advanced models into mass-market consumer products.
**Speed-Centric Consumer Market** This adjustment comes just four months after the model router launched with GPT-5. CEO Sam Altman initially positioned the feature as unifying user experience by eliminating confusing manual "model selectors." He noted the router increased free-tier usage of reasoning models from under 1% to 7%.
However, this costly bet on answer quality failed to gain expected traction. Sources indicate that while reasoning models deliver superior responses, most consumers prioritize speed over accuracy for casual chats.
Chris Clark, COO of AI inference provider OpenRouter, emphasized that response speed and tone are critical in general-purpose chatbots: "If users must watch a 'thinking' dot for 20 seconds after typing, engagement plummets." He noted these chatbots compete directly with Alphabet's search engine, which prioritizes instant results over "slow-but-precise" answers.
**Growth Concerns and Rising Competition** OpenAI's strategic shift coincides with intensifying competition. Altman recently declared a company-wide "Code Red" to refocus resources on core consumer products. Despite ChatGPT's 800M+ weekly active users, third-party data suggests slowing growth.
Traffic analytics firm SimilarWeb shows Gemini gaining momentum while ChatGPT's growth plateaus. Since September, average session durations on ChatGPT have fallen below Gemini's, according to SimilarWeb's VP of Data & DaaS Products.
Additionally, OpenAI CFO Sarah Friar reportedly told investors that user engagement dipped slightly after August's expanded content filters and safety classifiers—measures that inadvertently increased product friction.
**Cost Factors and Long-Term Bets** Beyond user experience, cost played a key role. Reasoning models are not only slower but significantly more expensive to operate. Redirecting free-tier queries to GPT-5.2 Instant helps optimize costs.
OpenAI noted GPT-5.2 Instant's improved performance now handles safety benchmarks better, eliminating the need to force-route sensitive conversations to reasoning models.
Despite this setback, OpenAI hasn't abandoned the concept entirely. Paying subscribers ($20/month Plus and $200/month Pro tiers) retain router access.
Industry experts maintain that dynamically allocating compute power based on query complexity remains the right long-term approach. "Fundamentally, using different models and compute for different problems makes sense," said Robert Nishihara, co-founder of AI platform Anyscale. "Regardless of short-term pivots, routing technology will persist." OpenAI plans to reintroduce the router for free and Go users after technical improvements, sources confirm.
Comments