I found that ~15% of my Astra requests are silently returning Luna in the model_id — while charging Astra rates

Overview

I have a $200/mo Pro 20x subscription. Recently Astra responses felt noticeably dumber on Codex, so I decided to inspect the packets. Turns out, ~15% of my gpt-6-astra requests on Codex are returning gpt-5.6-luna in the response body model_id . And my quota is still being consumed at Astra rates. This is on a $200/mo Pro 20x account. It gets worse. Another developer used F12 on the ChatGPT web app and found their gpt-5.6-sol requests returning gpt-5-5-mini in the model_slug field . So regular Ch

SpinGraph analysis pending — check back after processing.