Which variation is better: Caro Kann classical variation, Karpov variation or Tartakower variation?

I play the Tartakower variation

The Tartakower variation is more in fashion.

It's a matter of taste. All of them have different tabiyas with different plans.

The Classical 4... Bf5 is the most solid. When White castles long, Black can do the same but the positions leave White with a long-term space advantage on the kingside. More recently, Black castles short which is sharper and more dynamic.
The Karpov variation 4... Nd7 is more double edged than Bf5. In the Ng5 mainline, Black doesn't castle and plays with the king on f8.
The 4... Nf6 variation was pioneered by Nimzowitsch and after 5 Nxf6 branches into two depending on which way Black captures.
5... exf6 is now fashionable, Black castles short but can even push his pawns forward by h5-h4. It's not so easy to mobilize White's queenside pawn majority.
5... gxf6 is the sharpest (and riskiest) option after 3 Nc3 with Black castling long and using the open g-file to attack.

The real question for Caro-Kann devotees is how much time you should spend on these variations when you're more likely to face 3 e5.

Sorry to brag
But see I Destroyed the Caro kann

With 97%
Classical variation :D

@SANJAY-KURAMA2003 said ^

Sorry to brag
But see I Destroyed the Caro kann

https://lichess.org/gmzAtu0

With 97%
Classical variation :D

Your opponent destroyed himself.

The tartakower variation gives you the best chance tor a win. The rest are those "who remembers more theory" lines.

Tartakower