ChatGPT Pro Users Report Silent Downgrade to Mini Without Explanation
Key Takeaways
- ▸ChatGPT Pro users were systematically routed to mini models without notification, UI indication, or explanation
- ▸OpenAI support provided only generic responses over two months, with no investigation or acknowledgment of the issue
- ▸Affected sessions showed drastically reduced performance with disabled reasoning and instant thinking times
Summary
A ChatGPT Pro subscriber discovered that their $200/month subscription was being silently routed to the mini model instead of the full Pro model, with no UI indication or notification from OpenAI. The user analyzed chat metadata and found that affected sessions contained 'i-cot' or 'gpt-5-5-mini' in the resolved_model_slug property, confirmed through data exports and direct model identification when prompted. Other researchers and social media users reported similar issues over recent months, suggesting a broader problem affecting premium subscribers.
The affected sessions showed significantly degraded performance, with reasoning disabled, thinking times reduced by dozens of times, and the model identifying itself as 'GPT-5 mini' when questioned. Despite contacting OpenAI support for approximately two months, the user received only generic responses about model hallucinations and cache clearing, with no investigation, explanation, or resolution provided. The lack of transparency is particularly troubling because premium users have no visual cue or reliable way to verify they are receiving the service they paid for, raising concerns about OpenAI's accountability to paying customers.
- Multiple users reported similar issues, suggesting a systemic problem rather than isolated incident
- Paying customers have no transparent way to verify they're receiving their subscribed service tier
Editorial Opinion
This incident exposes a fundamental trust problem in premium AI services. OpenAI's silence—both in the product UI and through customer support—violates the basic transparency that customers paying $200/month for 'research-grade reasoning' deserve. Whether these downgrades are intentional capacity management or a technical failure, the lack of notification transforms the service from a premium offering into a misleading one. In an industry already scrutinized for AI safety and alignment, failing to communicate honestly about service degradation sets a troubling precedent.