Astra vs Opus 5.5 usage on the 20x plan, normalized for token efficiency
After normalizing for token efficiency, the author calculates Astra usage at only 52% of Opus 5.5 and says OpenAI needs to double usage and add a 5-hour limit to match Anthropic.
i calcualted current OpenAI usage vs Anthropic's on the 20x plan situation is extremely bad Astra sits only at 52% vs Opus 5.5 (numbers normalized for astra's token use efficiency vs opus i.e 4.33x) but as Tibo announced a poll today, if they 2x the usage, they'll even surpass Anthropic's current generous usage on Pro plans Astra will hit 104.75% vs Opus 5.5 (this is what we actually want) in my opinion, regardless of the poll result OpenAI has to incease Usage by 2x and introduce 5 hour limit if they really want to match Anthropic's usage quota on pro plans. and if they 2x the usage without a 5h limit, man, whole sentiment will flip in favor of OpenAI. (looks like OpenAI solved their compute at least, now we jsst need a very strong model from them.)
This post quotes another post: x.com/thsottiaux/status/2109152716954222849