A 2016 Nature study found the human eye can detect a single photon just above chance, while conscious sight needs about 5 to 9 photons.

My understanding of the physics is that there's no lower limit. People can detect even individual photons.

As long as the pulse has enough energy to change the state of the material in your retina, you will sense it. Intuitively, if an attosecond pulse can melt steel, you can bet you'll see "something" if you fire it at your eyes, even if it's the last thing you see.

• How much light is a photon, anyway? Does is take thousands or trillions of photons interacting with the eye to register as light?

Others mentioned there is a atudy showing it could be a single photon in an extreme controlled environment.

But in normal conditions, eyes chemically adjust to a reasonable sensitivity, our dynamic range is not limitless.
Eyes actually have the very quick contraction and expansion of the iris, and then that chemical tuning on top, which is why it can take some minutes to see well in strong darkness.

So taking something more realistic you would see in different cases: In a darkened hall you can make out a faint led at 40m distance, faint being about a μW, which is unusually dim for most leds you encounter. A μW might yield you 100μl (micro lumens).

xkcd puts a firefly as 600μl lumen, which is about a millionth of a dim light-bulb. That comes to the same brightness at 100m distance, which is reasonable for seeing fireflys well.

Both of these shine that light output on a sphere, which your eyes capture a small circle of, whose size is your maximally expanded pupil. Looking at wikipedia "the size of the pupil [varies] between 1.5 mm and 8 mm", which having to look at wikipedias source (thanks for nothing) tells me it is the diameter, so radius 4mm.
That equates to 2.5ppb of the total area for the 40m distant led, so 2.5 billionths. So of those 100μl, the eye receives 250fl (femto lumens).

Assuming the color averages to green, specifically 555nm green, that has 1/683 W/l. A photon has energy in Ws of c•h/555nm, which works out to 2.8 quintillion (555nm) photons per second per lumen.

250fl is 250000 quintillionths of a lumen, so we are talking 7000 photons a second.
Now in the dark you might notice your vision is quite blurry, and takes a while to pin stuff down, analogous to a longer exposure time. So to detect say blinking, you might need to wait 100 or 200ms. So maybe we can say you can comfortably see units 700 to 1400 photons in a sufficiently dark environment.

You can think of lightning: It is only visible for some 100s of microseconds (<0.5ms) but it is very well visible. Granted, a lot of photons are emitted in this very short though.

Whether a flash of light is visible comes down to a fundamental concept in visual neuroscience known as temporal summation or generally Bloch's Law.

Our eyes don't process light instantaneously, they integrate light over a short window of time (around 10 to 100 milliseconds, depending on whether you are looking through your central or peripheral vision).

Bloch's Law states that for very short durations, the perceived brightness of a flash depends on the total number of photons hitting the retina during that time window, not the duration itself.

So you can flash a standard light source like a normal LED very quickly, like at 1 microsecond or faster, it will be completely invisible. The flash was so short that the total number of photons entering your eye fell way below the minimum threshold required to trigger your photoreceptors and send a signal to your brain.

inversely you can increase the intensity of the light enough that it will always be seen regardless of the duration of the pulse, a high-powered laser pulsed for a femtosecond will remain visible.

Depends on how far away it is. 😌