While true, the latest opus model has 1m token context. Which is a lot more than the previous 200k limit. Hard to fill that up with regular work, but easy if you try to oneshot a whole product.