GPT-6 Astra Slammed by Users for Performance Deterioration
OpenAI's latest flagship model, GPT-6 Astra, has been criticized by users on X platform just one week after its release. The complaints claim that the model's performance has 'deteriorated' since its debut.
Developer synthwavedd posted a message saying 'Astra today feels significantly dumber,' joking that it was like undergoing a 'brain surgery' after its launch. Another developer, Pranjal Paliwal, who initially praised the model, checked Astra's code and found no evidence of AGI but noted a performance regression.
Some users, including Salio and researcher Md Ismail Sojal, experimented with comparing Astra's responses to the same prompts on its release day versus current results. They observed poorer quality answers from Astra. T3Chat founder Theo attributed the issues to increased instability compared to Claude Fable, suggesting that users may have been enjoying a 'honeymoon period' before experiencing worse results.
Interestingly, some users pointed out that OpenAI's previous flagship GPT-5.6 Sol experienced similar performance fluctuations in July, although OpenAI's executive Tibo Sottiaux denied intentionally weakening the model but acknowledged experimenting with 'reasoning effort' settings.