Brilliant move, why?

Well considering that you have only three queen moves and a4 will force a sacrifice after b5, then ... only qxb7 really competes.  Superficially, it looks like a free pawn, but it also looks like it could get your queen maybe trapped at some point after bd5.  And at best it looks greedy.  So I'd say nothing is brilliant about the queen move.  Who knows what the computers criteria are?  I'd just ignore the new fangled move names like "brilliant" and only pay attention to what is labelled blunders and maybe also mistakes, which are objective labels that go along with specific changes in the evaluation.  

Thank you. Do you know if these labels like "brilliant" "mistake" "blunder" etc is dependent on the level of the player? Maybe the computer said it was a "brilliant" move taking his level into consideration. It is not expected that a 1054 rating player can calculate what you just explained (at least i cannot), and thus it is considering a brilliant move maybe?

Inaccuracy, mistake and blunder are entirely objective, and correspond with the numerical difference in the engine's evaluation between the move you played and the best move.  I think mistake is up to a -/+1.0 difference and blunder is more than that.  Inaccuracy I'm not sure exactly, but its a minor mistake in the engine's view and I don't even look at those normally.  Those are the standard terms used by every engine.  "Brilliant" is a new fangled term that chess.com added, and I have no idea what the criteria are.  I just don't see how it could be useful given that I already look at everything labelled blunder and sometimes look at the mistakes.  And this proves it.  

If I had to guess, the engine is saying that qxp is only bad if you dive deep, so its brilliant to realize you should avoid that and pick the other move.  But I think a decent human can be highly suspicious of qxp even if they cant see all the lines, so yeah, the term brilliant just doesn't work in human terms.  

Qxb7 is answered by Bd5 which threatens both Ra7 trapping the queen and Bxg2. If Bxb4 then Rb8 Qxa6 Nxb4 and both the queen and g2 are under attack. So Qh5 is simply the only move and any strong player would play Qh5 by instinct in bullet. Obvious moves aren't brilliant.

whats this ??

Actually, four safe-looking queen moves (including 9. Qg5), of which 9. Qg5 actually hangs the queen (the bishop at d2 is pinned to the a5-e1 diagonal).

Basically brilliant is just the best move in the position according to a computer, that is "difficult to see". I have no idea how engines evaluate what is difficult to see and what is not, but considering that there are a lot of queen moves that don't instantly lose, yet have some serious drawbacks, computer considers playing the best move in such a situation "brilliant".

Most of my "brilliant" moves according to engine have been things like rook moves on backrank and "silent" king moves. If they happen to be the best moves in the position, they are often considered "brilliant" by the chess.com game analysis tool.

There seems to be a bug with "brilliant moves" at the moment. Look in the "Site Feedback" forum.

Infuriating that they dont explain what brilliant is. 

Guesswork: I will say that "Brilliant" is a move initially considered bad by the engine, but on further analysis proved to be much better.

We are all in search of the Kantian Sublime.