This variation is interesting because black can immediately challenge the knight and if white takes I think after ...gxf6 black gets alot of compensation for the slightly weakened pawn structure. the f6 pawn supports ..e5 and the open g file could give black very good attacking chances. black has equal development, equal central control (pawn on d4 versus e5), and the g file so I would probably castle queenside and the chances for each side to attack if white goes kingside would probably be favorable to black as he has the useful open g file while white has no half open files.

Edit: this is the position

Could u show us this so we can see?

nope....white play 6. c3 and his position is solid and is slight better.

korchnoi used to play this line for black, as well as other strong players, i play this line up to Nxf6, but  i take back with 5...exf6 its ok for black :-)

exf6 is more solid but seems to make it easier for white to dominate the center. White is hardly better in my opinion and black gets some chances.

The idea behind this line is that Black gets to develop freely with Bf5 or g4 , Bd6, Nd7, Qb6 or a5 , castle queenside and Rg8. Like somebody said, the pawn on f6 can support the e5 push if needed but it is not an absolute necessity all the time that black castle. The pawns on e7, f7, and f6 are difficult to penetrate and it is been my experience that the King is perfectly safe in the center for most of the game. 

Somebody claimed White should play 6. c3 and get an advantage but I do  not see why this is necessary or how this is playing for an advantage. d4 is not under any real threats and White would be happy to see e5 at this point because the Black king would  be too naked. White should play to develop his minor pieces and possibly try to force black to declare his kings intentions. c3 also weakens the Queenside pawn structure if White wants  to castle queenside. This is dangerous because with the g file open, I doubt White wants to castle into a firestorm.

That is my take on the position anyway. I am not claiming to be correct, but I do not mind playing this position as black at all. Also, I play the Caro-Kahn regularly and this is a line that I choose. I feel that it offers easier play for black than in some of the main line stuff.

It's called the Bronstein-Larsen Variation, after the two GMs who first played it.  Bronstein always played ...gxf6, the most aggressive choice - the doubled pawn comes to support the center and an eventual ...e5 push, while opening the g-file and complicating White's plans for his King.  Larsen also sometimes played ...exf6, using the e-file instead while maintaining a solid Kingside.

One advantage of the line is that most people have never seen it as White.  I've tried it in rated play four times, and none of my opponents played 5 Nxf6+ !!!  It was as if they were more afraid of a prepared opening line than of giving Black doubled pawns on move 5 . . .

I later found the best counter to this "declining" line:  5 Ne4-g3 h5  6 h4  Bg4  7 f3 Bc8! when White has apparently only wasted time and weakened himself.

I forgot about this line. I think it's pretty interesting as with the g file, ...e5 and ...f5 possibilities, he has more chances for aggressive play than usual. I might start playing the ck for this line. I realize now that black can't really afford to immeiately play ...e5 and should just settle for ...e6, but there is still the possibility.

6. c3 is regarded as the best move in the diagram position because it more or less forces black to make up his mind where he will place his pieces.
It certainly is the best way if white is going to follow up with a fianchetto.

See, 6. g3 is inaccurate because of ..Qd5, and 6. Nf3 allows the pin Bg4. 6. c3 is in fact a good waiting move. Since, just as you said, black most certainly is going to castle queenside, it is also a good preparatory move for b4 and a further advance of the b-pawn, opening lines against the black king. a bishop on g2 will also be aiming at the black king.

That's pretty much why 6. c3 is regarded as the antidote to the Bronstein variation.

I think it's a waiting move with the idea that c3 will be played anyway.

I see your point...where does the white king want to go?

White does castle kingside. He plays g3 to blunt the activity of the g file. Black has a weakness on c6, the bishop aims there, and he has more space so he definitley will have an attack also.

As a Nc3 player against the Caro, I have always found Nf6 the most difficult continuation to play against.  So what do you recommend as the antidote?

Simply go into Nxf6+ gxf6 c3 followed by g3 with a slight edge because white has more space, can somewhat blunt the g file, and no weaknesses and that g2 bishop and c6 pawn will be fun for ramming the white b pawn against. Plus white may get a slight edge in the endgame and eventually try to exploit those doubled pawns. It's an interesting position with counterplay for black.

And in case of the exf6 variation, I'd recommend playing a timid setup with Nf3, Be2, O-O etc, and don't play too aggressively, but try to exchange pieces. With the pawn structure after exf6, all endgames are very favorable for white.