🚨 I can confirm this. GPT-Astra Medium burns MORE usage than xHigh. 👀
I tried both myself and the difference is wild.
Medium: usage disappears fast.
xHigh: somehow lasts WAY longer.
We were told to use Low/Medium to save usage… 😭
Turns out xHigh and higher reasoning might actually be the better choice.
what the hell is going on here? 👀