Read that http://jmonkeyengine.com/blog/
just install JMP alpha2 and i'm pretty sure it will help you to make your decision 
If you were already used to JME2, and that you had a specific and short scheduled project, i'd say JME2…
If you are starting from scratch, go for JME3.
nehon said:
If you were already used to JME2, and that you had a specific and short scheduled project, i'd say JME2...
If you are starting from scratch, go for JME3.
How would you compare the learningcurves ? I am not scared to learn, I just wanna know what kinda timeframes each are looking at, just a hint.
Because I am porting over code that's working fine I am eager to get it up running again, but in this awesome jME! :)
I can't compare, because i never used JME2.
But about the learning curve : i learned about JME3 in the end of may, and in one month i've been able to load a terrain, put a character in it, apply physics, play animations, make my character walk on the terrain, and…learned to use blender in the process 
… I had some basic knowledges about 3D though, and i'm a Java programmer for a living.
The result is just 3 or 4 very concise Java classes and a couple of assets.
Besides, the wiki documentation is growing everyday.
From what are you porting code?
I have been using JME3 since July more or less. I played a bit with JME2 back in the first months of 2009 but never got too far (I liked a lot JME2 but I had a shift of interests by then).
About JME3, the first thing I noticed when I started using it was that its general architecture felt familiar and easy to grasp for me (I am a seasoned developer, last ten years mainly Java. But I started with a spectrum 16kb).
The second thing is that I come from the IT business world, and I am a beginner regarding computer graphics (currently studying a lot about it in my spare time, and getting very excited by how much more I understand today than I did a few months ago :)). In this sense, the JME3 SDK Platform has helped me to not get lost, by providing a way to organize the project and sorting the concepts in my head by the way.
Last but not least, my intention is to make intensive use of procedural techniques to generate some of the gameworld contents. When I saw the elegant Materials system and the support for shaders I just loved it :). Until now all I am doing is learning and trying things that I want to have ingame (such as planetary atmosphere scattering, with which I am struggling for two weeks now, which is a shame because all I am doing is porting Sean O'Neil's work into JME3…, so far I got half of it working).
These are the main three reasons why I chose JME3.
Cheers.
@Twixly maybe if you go into some technical detail about the core mechanics/features of your game (e.g. what type of graphics you rely on or desire) it'll be easier for the developers to assess your needs.
About learning curves, I'd say we haven't done our job right if jME3 isn't as easy or even easier to catch on to than jME2
Thanks for all the feedback!
I've decided to go with jME3. We are going for graphics kinda like the one in "Nord" which is in your movies and demo section. Kinda cartoony cute.
Once we get stuff working I'll give some links etc somewhere on the forum.
Oh, and we used GTGE and was doing pure 2d isometric, however due to the limitations we've run into we decided an open 3d view would serve us better. Thanks.