Somewhat related, the newest sensor capable in resolving photons is a stacked BSI sensor.
https://www.eetasia.com/meet-the-first-quantum-image-sensor/0.3e- read noise, "detecting every photon" is an exaggerated (well, false) claim but >90% is already exceptional.
To put this into perspective, a sensor with 0.5e-rms read noise, which is a less than double of 0.3 will misclassify ~32% photons, making it incapable of photon number resolution.
Calculations are a bit more involved, we can assume photoelectron distribution is Poisson, and the statistical uncertainty (read noise!) is Gaussian.
To calculate the amount of photons resolved, one would use the Gaussian cumulative probability function to find the area under the curve, which is done through integrating [-e-rms, +e-rms] and then subtracting the value from 1.
https://en.wikipedia.org/wiki/Normal_distribution Under "CDF".
A read noise of 0 is needed to resolve every photon, which is impossible. I suppose semantics is involved, detecting a photon is a little different to being able to
detect and resolve it.
There is one commercially available camera that does this:
https://www.hamamatsu.com/jp/en/product/cameras/qcmos-cameras/index.html Pretty much a "M 3/4" sized sensor.