Tell HN: OpenAI has been silently routing GPT-5.3-Codex requests to GPT-5.2 (github.com)

Users of the GPT-5.3-Codex are reporting a puzzling issue where their configurations intended for the latest model are being silently routed to an earlier version, GPT-5.2. This unexpected behavior has sparked frustration among subscribers, who are experiencing productivity losses as they unknowingly work with outputs from the older model while believing they are using the newest version. Users have documented the anomaly through detailed logs showing mismatched model responses, highlighting a critical flaw in communication regarding model routing. This incident raises significant concerns within the AI/ML community about transparency and user agency. The ability to detect which model a user is interacting with is essential for ensuring accuracy and accountability in AI outputs, especially in professional settings where decisions are being made based on these results. The situation not only calls into question the reliability of the versioning system but also emphasizes the need for improved notification mechanisms when users are switched between models. As AI tools increasingly integrate into workflows, ensuring users are aware of what system they are accessing becomes paramount to maintaining trust and efficiency in AI-assisted tasks.