we dont' need to "freeze" models if next versions aren't a huge improvement over previous ones. That seems to be case now.

Really? Only last month OpenAI declared the "AGI era" is here.