It’s exciting that a model scoring this high is dirt cheap.

It’s also so inefficient, when they release the full performance numbers it’s not going to be good.

One example, it takes about 3.6x more tokens to finish the same work as Gemini Flash 3.6.