Previous ARG OH poll was +1 Obama so that's a 2 point swing, not sure about Colorado
ARG really doesn't have a consistent lean, if you look at 538 -- they're always somewhere and it's never near the consensus. The model does not give them a ton of respect.
Nate Cohn has a timely article on Gallup's likely voter model:
This afternoon, Gallup applied its likely voter model for the first time and found Romney gaining 5 points compared to registered voters. But while a persistent likely-registered voter gap isn't good for the president's chances, it's worth remembering that this is about the same difference that Gallup found at this point four years ago. As you can see, Gallup initially showed a 6 point gap between likely and registered voters, which shrunk to just a couple of points by Election Day.
http://www.tnr.com/blog/electionate/108342/romneys-gain-among-likely-voters-not-unusual