Check is your OpenAI GPT-6-Astra nerfed? (is-my-astra-real.pages.dev)

Recent investigations have revealed that some users of OpenAI's GPT-6-Astra may have been experiencing a downgraded performance, akin to earlier GPT-5.6-Luna outputs. The analysis, which involved running the same prompts repeatedly, indicated that around half of the models drawn by affected accounts produced crude cartoonish versions of requested images, with the affected instances being statistically slower and less coherent than their Astra counterparts. This issue seems to have emerged abruptly on September 20, 2026, and is believed to relate to account-specific or routing anomalies. The significance of this development lies in its implications for users relying on consistent performance for tasks like content creation and data analysis. Affected accounts showed around 25% less output quality, with more interruptions from users indicating dissatisfaction with the responses. While differences were significant, the exact cause remains unclear, leaving users in the dark about whether they are genuinely using the advanced model. To check their performance, users can utilize a Python script that compares their account's output against expected standards. This situation raises important questions about OpenAI's model stability and the transparency of its AI deployment policies.