Heavy Users Complain OpenAI's New Models Lost Intent Understanding
Heavy ChatGPT users, including one burning over 3 billion tokens monthly, complain that GPT-6.2 Sol and Astra follow instructions too literally while losing grasp of user intent, with some preferring the older Fable 5.2 model.
2026-11-21 ~ 2026-11-22 · 2 related posts
- Heavy user: GPT-6.2 Sol and Astra follow instructions better but lost implicit intent understanding — that_antimatter · 2026-11-21
- User says newer OpenAI model misses intent, claims older one 'gets you' — imjustnewatai · 2026-11-22