Elon's predictions for Grok read very honestly and extremely realistically.

Grok 4.7, which is coming soon, will be on par with Opus 5, in some aspects better, in others worse.

Grok 4.8, which is a 2.5T model trained with their new C++ software stack, will be a significant leap, and the subsequent models will presumably be on par with Fable.

Considering the recent releases, these are, in my opinion, truly realistic goals, neither too ambitious nor too low. SpaceXai was the last to enter the race but has recently caught up considerably with Cursor and enabled some truly excellent releases. I'm really looking forward to the next models and think that the competition is the best thing that could have happened to us.