From Hecht, S., Shlaer, S., & Pirenne, M. H. (1942). Energy, quanta, and vision. The Journal of general physiology, 25(6), 819-840.:
Direct measurements of the minimum energy required for threshold vision
under optimal physiological conditions yield values between 2.1 and 5.7 X
10-10 ergs at the cornea, which correspond to between 54 and 148 quanta of
in order to produce a visual effect, one quantum must be
absorbed by each of 5 to 14 rods in the retina
They're talking about blue-green light, which rods are most sensitive to. I agree with you that light intensity in units of power density doesn't seem quite right, I'd think about it in terms of counting photons instead. Basically they estimated you need somewhere around 10 photons absorbed, with about 1/10 photons being detected. A more recent paper suggests humans can detect as little as a single photon:
Tinsley, J. N., Molodtsov, M. I., Prevedel, R., Wartmann, D., Espigulé-Pons, J., Lauwers, M., & Vaziri, A. (2016). Direct detection of a single photon by humans. Nature communications, 7(1), 1-9.
As far as "red", it'll depend how "red" you go. Rods aren't particularly sensitive to deep red light (I see quite a range of estimates depending on experimental conditions, but e.g. Lamb, T. D. (1995). Photoreceptor spectral sensitivities: common shape in the long-wavelength region. Vision research, 35(22), 3083-3091. shows about 4 to 510 less sensitivity of human rods at 700 nm compared to peak), and cones are far less sensitive to light because they require multiple photon hits on a single cone to detect; they're also sparse in the periphery. See for example:
Donner, K. (1992). Noise and the absolute thresholds of cone and rod vision. Vision research, 32(5), 853-866.
As such, when people are interested in the thresholds of human vision, they're not usually looking in the red spectrum.