with almost no impact by Stein (1%) or Johnson (2%). Details are here
However, I think their model, like that earlier today of Quinnipiac, may have some problems.
For starters, their sample is 80% white, with Trump winning those white voters47-41.
Ballotpedia had a Florida sample that was 64% white, while the actual percentage as seen in CNN exit polls was 67% white (Romney winning 61-37). Worth noting that Trump is badly underperforming Romney among White voters in a state that Romney lost.
They show an 11% Latino, split 5% Cuban and 6% other. The Cubans break for Trump 45-34 while the non-Cuban breaks for Clinton 61-26. Ballotpedia has Florida 17% Hispanic with no breakout between Cuban and non-Cuban, while the actual exits showed (Obama winning 61-38 — remember that Cubans tend to vote much more Republican).
The sample is also 13% African American, with Clinton winning these 76-10. Ballotpedia has a model that is 16% Black, while the 2012 exits were 13% African American (Clinton winning 95%).
2015 Quick facts from the Census Bureau says the state was 77.7% White only (up from 75% in 2010), 16.8% African-American (up from 16.0), and 24.5% Hispanic, up from 22.5).
In other words, while the percentage of African-Americans seems right, the percentages of whites seems overstated and of Hispanics understated. Despite that, Clinton shows a lead of 4%, while in 2012 Obama only carried the state over Romney by 50.1% to 49.13.
In short, despite the top line being relatively narrow, and even given what I think is a fairly incorrect demographic model, Clinton is performing better than Obama did when carrying the state in 2012. It is very hard to imagine a scenario of a Republican winning the Presidency at this point if s/he loses Virginia.