What each board ranks

The four block.space rankings, the leaderboard, largest inscriptions, heaviest tokens and collections: what one row is, which book each reads by default, how deep it goes, and the rules they share.

block.space's rankings page, block.space/rankings, gathers four boards under one heading: "one measure: the blockspace bought". Each ranks by bytes, never by count, and each reads the same books under a different question. The book names are explained on the three books.

At a glance #

board	one row is	default measure	depth
leaderboard	one address	Book A, divided among holders	every address the appendix can place
largest inscriptions	one inscription	Book C, the file's own bytes	top 100,000 per family
heaviest tokens	one token, any protocol	Book A, everything it bought	top 100,000 per family and overall
collections	one NFT collection	Book A, its members' books summed	top 10,000 per book, unit and family

Every board has the same toggles: Book A, B or C, and bytes or vbytes.

The leaderboard #

An address's figure is "the blockspace behind the tokens it holds: for each token, the token's bytes under the chosen book, divided among its holders by their share of the supply". Holding half of a token's units means half its bytes. The share is by units held, not by value, and a non-fungible unit such as an inscribed sat has one unit, so its holder carries all of that token's bytes. The share is taken of the holder total the source reports, not of the protocol's stated supply.

A token whose holders no source can see adds nothing to any address. The board's coverage line says how much of the books that leaves out; see coverage. On the latest reading the leaderboard ranked 4,491,928 addresses (as of block 972,261, 2026-10-20 11:17:25 UTC). Holders are keyed by script, and the address is a rendering of it; non-standard scripts are shown by hash.

The largest inscriptions #

An inscription's size is Book C, "the bytes of the file itself, with every byte of framing removed". The board ranks bytes bought, not bytes shown: a delegate inscription that displays another inscription's image "bought a few hundred bytes and is that size". One row is one inscription. A "by sat" view ranks the sats instead, so a sat inscribed many times carries all its inscriptions.

Tabs rank other content families by the same book: Counterparty assets (which carry Stamps and Counters), SRC-20, BRC-20 and TAP. For BRC-20 and TAP, a token's content is the sum of many small files, and the board says so.

On a read of the rankings API on 2026-10-20, the top row was inscription #79,351,656, a text/html file of 3,969,429 bytes at block 873,893.

The heaviest tokens #

A token's weight is Book A: "every byte of every transaction its operations bought, shared where shared". Under the Shared Claim Axiom a byte two tokens bought is divided between them, never counted twice. A sat, a rune, a BRC-20 tick and a Counterparty asset stand in one ranking "because the measure is bytes, not kind". Each row carries the fees its transactions paid, divided by the same shares.

On the same read, the top three were UNCOMMON•GOODS (Runes, 23.9 GB), DIESEL (Alkanes, 11.0 GB) and Tether (Omni, 6.50 GB).

Collections #

A collection is a set of NFTs with a named source of membership, and each row shows its provenance:

chain: an inscription's children, and SRC-101 roots;
index: Counterparty subasset families and Atomicals containers;
asserted: a published list, named with its source, never merged with a chain group.

Fungible tokens are left out even where they are grouped. A collection's bytes are the sum of its members' books. On the 2026-10-20 read the largest by members was the children of inscription #66,166,842, with 210,007 members, and the page said its rows were "ranked by members while the bytes resolve".

From block.space's rankings documentation:

Every figure is bounded by a height: "as of block H", the block the holders appendix is synchronised to. See the frontier and "as of".
Nothing is sorted at request time; each board reads a ranking written in advance.
The population is stated, with the cut-off. A figure below it is "unranked, not zero".
"Not measured" is never 0. See null is never zero.
Labels come from the ordinal record; numbers come from the books. Where they disagree, both are shown.
Colour names the family, never the rank, and the key is the same on every page.
The API carries the exact, unrounded values under one envelope: the heights, the mode, the population, the total and the rows.

Sources #

Live figures: holders as of block 972,261, 2026-10-20 11:17:25 UTC. Refreshed hourly.