This site holds 304 pullable characters across four games, out of 328 rows in total. Putting the four side by side is the obvious thing to do with that, and it goes wrong quickly, because the four games do not agree on what they are counting.
The headcounts are easy enough. Genshin has 118 pullable characters, Star Rail 87, Zenless Zone Zero 60 and Honkai Impact 3rd 39. Genshin has been running longest and has the largest cast; nothing surprising is happening there.
Rarity is not a shared scale
Star Rail has 74% five-star characters and Genshin has 57%. Zenless Zone Zero instead uses native ranks: 13 A-rank, 47 S-rank. An S-rank agent is not a five-star character, and converting those ranks into stars would invent a shared scale. These distributions describe each roster; they do not measure which game is more generous.
Honkai Impact 3rd does not appear in that comparison at all, and the reason is worth dwelling on. It has no rarity system for characters. Its share is not 0% and it is not low — it does not exist. Everywhere on this site that would print a percentage for it prints nothing instead, because a 0 in that column would say “no five-stars,” which is a claim about Honkai Impact 3rd rather than a claim about our data.
That distinction runs through the whole site. A character with no votes is absent from a leaderboard rather than sitting at zero, for the same reason: zero means nobody would pull for them, which is the opposite of not knowing.
Honkai Impact 3rd counts something else entirely
Honkai Impact 3rd’s 39 entries are characters, but what you actually obtain in that game is a battlesuit, and one character can have many. Across those 39 rows there are 110 battlesuits. Counted that way it is not the smallest roster here by a wide margin; counted by character it is.
Neither count is a trick. They answer different questions, and the site had to pick one because the game asks “would you pull for this?” about one thing at a time. It picked the character, which makes Honkai Impact 3rd look small in every chart above, and that is a limitation of the comparison rather than a finding about the game.
What is not comparable at all
Release dates are the sharpest example. The roster carries them for every Genshin character and for none of the other three — the upstream sources this site rebuilds from simply do not publish them in a form worth trusting. So the Genshin timeline post exists and the other three do not, and no amount of wanting the comparison makes the data appear.
The same is true of taxonomy. All four games sort characters into something like an element and something like a role, but the sets are different sizes, they mean different things, and one of them has neither. Only 3 of the four have rarity. Comparing the categories directly would require flattening distinctions the games themselves treat as central, which is why every chart here stays inside one game unless the quantity genuinely survives crossing between them.
What does survive is the question this site asks, which needs no taxonomy at all: shown one character, would you spend money on that banner? That has the same meaning in all four games, and it is the only column where the four rosters can honestly sit in one table.