Because 32 bit/channel is floating point, and gamma 1.0.
Your 8 bit values are probably gamma 2.2 (or whatever colorspace you are using).
8 bits and 16 bits images have a gamma 1.0, this is in 32 bits that the gamma is higher.
How do i set gamma = 1 for 32 bits images? I want 8, 16 and 32 bits images behave the same.