Key Points
- OpenAI removes the automated model router for Free and $5‑a‑month Go users.
- Default chat for these tiers now uses GPT‑5.2 Instant, the fastest and cheapest model.
- Users can still manually select advanced reasoning models if desired.
- The router stays active for paid tiers ($20‑a‑month Plus, $200‑a‑month Pro).
- User feedback showed a preference for the simpler default experience.
- Router usage among free users rose from less than 1% to 7%, increasing costs.
- Competition from Google Gemini is intensifying, with visit duration shifting.
- Safety improvements in GPT‑5.2 Instant reduce the need for routing sensitive queries.
- OpenAI may reintroduce the router for free users after further refinement.
SAN FRANCISCO CALIFORNIA JUNE 02 Open AI CEO Sam Altman speaks during Snowflake Summit 2025 at Moscone Center on June…
Background
OpenAI introduced an automated model router that evaluates user queries and directs them to either a fast, low‑cost model or a slower, more powerful reasoning model. The router was intended to streamline the experience by automatically selecting the most suitable model for each question.
User Feedback and Rollback
After monitoring usage, OpenAI found that the router increased the share of reasoning‑model usage among free users from less than 1 percent to 7 percent, a shift that raised operational costs. Feedback indicated that many free users preferred staying in the default chat experience and did not want the router’s automatic selections. As a result, OpenAI announced that free and Go users will now default to GPT‑5.2 Instant, the fastest and cheapest model, while still retaining the option to manually choose a reasoning model when desired.
Impact on Paid Subscribers
The model router will remain active for paid tiers, including the $20‑a‑month Plus plan and the $200‑a‑month Pro plan. These users continue to benefit from the router’s ability to route complex queries to advanced models, reflecting a different usage pattern and willingness to pay for higher‑performance responses.
Competitive Landscape
The rollback occurs as OpenAI faces intensified competition, particularly from Google’s Gemini service, which has shown growth in recent months. Industry data suggests that average visit duration on ChatGPT has fallen below that of Gemini, highlighting the pressure on OpenAI to maintain user engagement.
Safety and Performance Enhancements
OpenAI also noted that the newer GPT‑5.2 Instant model has improved safety benchmark performance, reducing the need to route sensitive conversations to higher‑cost reasoning models. The company continues to implement safeguards, such as prompting users to take breaks during long sessions and expanding content classifiers to block unsafe material.
Future Outlook
OpenAI indicated that the model router may be relaunched for free users once it is refined, signaling an ongoing commitment to balancing user experience, cost efficiency, and safety. Experts predict that model routing will remain a key strategy for matching computational resources to problem complexity over the long term.
Source: wired.com