Why shouldn’t they be? They are probably the same size and cost the same to run. They are not doing full training runs (eg Mythos) so don’t need to recover insane training costs.
replyI'd be kind of shocked if a model that came out six months ago is the same size and cost to run as one that just came out today.
replySame size? Maybe by a bit. Cost? Absolutely. Newer flagship models are often slightly larger each generation, but not even 2x. But more efficient architectures are coming out all the time, and it'd be a waste to retrain an old model. So it washes out.
replyOpus 4.7 and presumably 4.8 are more expensive due to a new tokenizer that translates data into more tokens per input.
replySame price on a token basis, but usually steadily decreasing on a task basis
replyDidn't you mean increasing?
reply