The open loop response I'm getting from the ADA4895 SPICE model does not look very realistic when compared to the datasheet. Besides the dominant pole all I see is a single pole-zero pair around 1 GHz giving a phase dip down to about -145 degrees before returning to -90 deg at higher frequencies. I was thinking the model would be fairly detailed since the netlist includes a bunch of additional poles and zeroes but on closer inspection it looks like the three additional poles and three additional zeros are all just stacked up at 1 THz and have no effect on the response. Is that intentional or did something go wrong when generating this model?
The revised SPICE model for the ADA4895 is attached. This should have the noise modeled correctly.