Also is rolling out to the $100/$200 Codex subscriptions, I got access this morning.
Unsurprisingly, it does indeed consume usage at ~2.5x the rate of Sol.
pllbnk•4m ago
Shouldn't this new reduction in thinking output make the model cheaper to operate? Could it be that the model is the same old LLM, with a bit newer architecture but still doing the same things, including huge amounts of thinking, just that its thinking is less clear to human readers?
minimaxir•33m ago
Unsurprisingly, it does indeed consume usage at ~2.5x the rate of Sol.
pllbnk•4m ago