There's only been a few polls updated since the last run of the SanityIID Senate model. The changes those few polls made only had the effect of making the Republicans slightly more likely to retake the Senate.
As for the details; Alaska, Colorado, and Iowa are a tossup. Their mean (average) polling is so close that statistics aren't really meaningful. They are essentially a coin toss at this point in time. The Monte Carlo distribution I am using shows this in the elevated probabilities of 48, 49, or 50 seats.
As for Kansas, it desperately needs to clarify the issue of exactly which candidates are going to be on the ballot. Once that is done, there needs to be more polling if there is going to be better predictions.
As of this writing, my model is showing Kansas with about a 75% chance of electing Orman, but I also believe these percentages to be overly skewed toward him due to lack of polling. There are only four (4) polls that come from different polling places and are relatively new. Since the state of Kansas is still in flux, expect this to change in the coming weeks and months.