**I'm assuming you mean 'how small' could they be?**

For the viability of a humanoid body plan we can look to the [pigmy marmoset][1].

Smaller than that may be possible but clearly 'as small as' definitely is.

[![enter image description here][2]][2]

Could something that small speak? yes, ask any budgerigar.

Can they fly at that size? lots of things that size fly.

It's just a matter of giving them adequate wings & appropriate musculature to drive them.

The smallest mammal by weight (I assume you want them to be mammals?) is the [Etruscan shrew][3] weighing about 1.8 g (0.063 oz) on average with a body length of about 4 cm (1.6 in).

[![enter image description here][4]][4]

The [bumble bee bat][5] (a flying one for you) is about 29 to 33 mm (1.1 to 1.3 in) long & 2 g (0.071 oz).

[![enter image description here][6]][6]

So I see no reason you couldn't go down to their size (though as there's nothing smaller than the Etruscan shrew that's likely your smallest reasonable limit for a mammal).


Their voices may be a bit distorted & high pitched that small but I see no reason they couldn't still talk.

How would their biology & diet differ? shorter lifespans, higher metabolism, faster heartrate & having to feed far more frequently are strongly indicated by our knowledge of small mammals.

*Intelligence is a whole other ball game, human intelligence in a hard science context for something this small is not really possible, with a bit of handwaving you might get away with equivalence to [a Three year old  toddler][7] for a larger (than a smallest possible) model without too much grumbling from the pedants, but the smaller you go the softer the science & ever more handwaving needed I fear.*

-----

Have a look at [this answer][8] (you've reminded me I haven't finished tidying it up, thanks, I'll have to get onto that) to a related question & other answers to it, you may find they give you some ideas.

-----

*Later edit: I'm 100% satisfied with my conclusions on size but not entirely with those on intelligence.*

*In tests a Raven does roughly as well as a Chimp & both are considered roughly equal to a two year old, they do better in some areas but broadly speaking across all tests that seems to be the consensus on where they average out.*

*But the Raven does it with more or less 4% of a Chimps brain mass (about 15g), we think they achieve this with a greater neuron density than in mammals.*

*Extrapolation from that suggests a 54g brain (assuming the human average of 2% of body mass that's somewhere between a one foot six inch & two foot tall person I think? meh! someone else will have to do the math on that for you) using the same tricks birds do to cut brain weight (to make flight easier) could plausibly be as bright as an adult human.*

*But I'm far from convinced it can simply be scaled like that to produce human level intelligence & the danger is new work in the field could potentially make any such extrapolation appear silly tomorrow or that there already is work out there that does I've just not come across.*

*You need an expert in the area for that one.*


  [1]: https://en.wikipedia.org/wiki/Pygmy_marmoset
  [2]: https://i.sstatic.net/t3THC.png
  [3]: https://en.wikipedia.org/wiki/Etruscan_shrew
  [4]: https://i.sstatic.net/BBU0a.png
  [5]: https://en.wikipedia.org/wiki/Kitti%27s_hog-nosed_bat
  [6]: https://i.sstatic.net/p4mcM.png
  [7]: https://raisingchildren.net.au/toddlers/development/development-tracker-1-3-years/2-3-years
  [8]: https://worldbuilding.stackexchange.com/a/158515/49261