Ki, or affinity, is determined by an equation that basically considers
1. how quickly a ligand binds to a receptor
2. how long it stays there
So if a ligand binds quickly and then stays a while once bound, it has high affinity. If it binds slowly and unbinds quickly, then it would have low affinity and wouldn't be very good at activating the receptor.
Drugs can interact with the receptors in different ways, for example very high affinity ligands can form covalent bonds with receptors and stay bound almost indefinitely (until the receptor complex recycles). But mostly they're just interacting with residues, however I don't know too much about chemistry.
So there may be some spots on the receptors that the ligands fit neatly into (like a key in a lock) and then some interactions cause the receptor to change shape a bit (a conformational change) which then produces some change in the receptors, possibly activating it all the way like the endogenous ligand (full agonism), activating it part of the way (partial agonism), or not activating it at all (antagonism).
But ligands can have high affinity for a receptor and just sit on top of it, not actually activating it (the ability to activate a receptor takes some intrinsic efficacy). So the affinity is how attracted a drug is to the receptor, while the intrinsic efficacy is how good the drug is at actually activating the receptor once its bound.
You can have a drug that is great at binding to receptors but doesn't activate them (a high affinity antagonist) and a drug that is great at activating receptors but hardly ever binds (a low affinity agonist).
As a caveat, benzos aren't really agonists IIRC, they are known as positive allosteric modulators. They function by binding to a spot on the GABA-A complex that, when the benzos bind there, increases the affinity of endogenous GABA for the GABA-A receptor.
So benzos work by enhancing GABA's affinity, if you will. At least I don't think the benzos have much of their own intrinsic efficacy (in other words, benzos may not actually do much in the absence of endogenous GABA). Whereas I believe barbiturates have some intrinsic efficacy.
Hope this was helpful