ADA4896 LTspice model noise vs temperature

When simulating voltage noise versus temperature of an ADA4896 in follower configuration:

I observe a huge increase in voltage noise spectral density ranging from 0.95 nV/rtHz at 25 °C to 8.9 nV/rtHz at 105 °C (at 10 kHz).

This seems unrealistic, could you comment on that ?