Skip to content

Lotus Root v12.4.16

Choose a tag to compare

@Boblepointu Boblepointu released this 17 Nov 09:51
· 2 commits to main since this release

Pool links that no longer stall, faster block connection, Taproot addresses, one GPU miner

Important

Napoleon, 2026-12-21 20:50 UTC: every node must run 12.4.13 or later by then. This version qualifies.

Warning

GPU miner users: the lotus-gpu-miner-amd image and the -nvidia / -amd archives are no longer produced. Use lotus-gpu-miner for both vendors, and pass your own payout address: there is no built-in one any more. AMD cards were not tested on real hardware for this release.

  • Pool links. Two pool nodes that connect no longer stop reading each other while both send their catch-up; such links used to stay dead until the ping timeout.
  • Faster block connection. The explorer and module databases are written after the block, in order, on their own thread: connecting a large block takes about half the time. -asyncindex=0 keeps the old path.
  • Batch submission and light templates. sendrawtransactions admits up to 1,000 transactions per call, 2.2 to 2.7 times faster than separate calls; getblocktemplatelight answers in under a millisecond.
  • Taproot addresses. The node wallet, the web wallet, the miners and the pool can create, receive on and spend from Taproot (lotus_J…) addresses. Legacy addresses stay the default.
  • One GPU miner. A single program and image for NVIDIA and AMD, with --backend auto|nvidia|amd.

New here? The install and upgrade guide takes you from download to a running node.

Downloads

Node and tools

The archive holds lotusd and its tools (lotus-cli, lotus-tx, lotus-wallet, and lotus-seeder where it is built).

Platform File Size
Linux x86_64 lotus-binaries-12.4.16-linux-x86_64.tar.gz
one program only: lotusd · lotus-cli · lotus-tx · lotus-wallet · lotus-seeder
32.8 MB
Windows x86_64 lotus-binaries-12.4.16-windows-x86_64.zip
one program only: lotusd · lotus-cli · lotus-tx · lotus-wallet
17.9 MB
macOS Intel (x86_64) lotus-binaries-12.4.16-macos-x86_64.tar.gz
one program only: lotusd · lotus-cli · lotus-tx · lotus-wallet · lotus-seeder
29.3 MB

GPU miner

Platform File Size
Linux x86_64 lotus-gpu-miner-12.4.16-linux-x86_64.tar.gz 2.7 MB

Lotus Miner desktop app

Platform File Size
Linux x86_64 lotus-miner-gui-12.4.16-linux-x86_64.tar.gz 20.9 MB
Windows x86_64 lotus-miner-gui-12.4.16-windows-x86_64.zip 10.3 MB
macOS Intel (x86_64) lotus-miner-gui-12.4.16-macos-x86_64.zip 9.8 MB
macOS Apple Silicon (arm64) lotus-miner-gui-12.4.16-macos-arm64.zip 8.9 MB
macOS universal (Intel + Apple Silicon) lotus-miner-gui-12.4.16-macos-universal.zip 18.3 MB

Checksums of every file: SHA256SUMS.

Docker

docker pull ghcr.io/lotusdevorg/lotusd:12.4.16            # the node
docker pull ghcr.io/lotusdevorg/lotus-cli:12.4.16         # command-line client
docker pull ghcr.io/lotusdevorg/lotus-tx:12.4.16          # transaction tool
docker pull ghcr.io/lotusdevorg/lotus-seeder:12.4.16      # DNS seeder
docker pull ghcr.io/lotusdevorg/lotus-wallet:12.4.16      # wallet tool
docker pull ghcr.io/lotusdevorg/lotus-gpu-miner:12.4.16   # GPU miner

To pin the exact node image, pull it by digest:

docker pull ghcr.io/lotusdevorg/lotusd@sha256:f3bcd9ab77ee0100662d297a23e394571b61e04b9d1e36ba6db2014d62c19d92

Verify your download

Every file of this release is listed in SHA256SUMS. Download it next to your file and check:

curl -fsSLO https://github.com/LotusDevOrg/lotusd-releases/releases/download/v12.4.16/SHA256SUMS
curl -fsSLO https://github.com/LotusDevOrg/lotusd-releases/releases/download/v12.4.16/lotus-binaries-12.4.16-linux-x86_64.tar.gz
sha256sum -c --ignore-missing SHA256SUMS     # Linux; must print: OK
shasum -a 256 -c --ignore-missing SHA256SUMS # macOS

On Windows, Get-FileHash <file> -Algorithm SHA256 in PowerShell must give the value on that file's line. The files are not signed: download them from this page only.

Full release notes

12.4.16 is 12.4.15 plus a fix for pool links that stalled right after
connecting, an asynchronous explorer and module indexer that takes most of
the work out of block connection, batch transaction submission and light
block templates, faster block storage and UTXO set hashing, a thin share relay, mining views by scope and unified
documentation in the node web UI, Taproot (P2TR) addresses in the wallets, the miners and
the pool, and one GPU miner for NVIDIA and AMD. No
consensus change
: the Napoleon rules and their activation (median time
past 1797886200, 2026-12-21 20:50:00 UTC) are those of 12.4.13 to 12.4.15,
and Taproot is valid on mainnet since Second Samuel (median time past
1766329380, 2025-12-21, block 1189225). A 12.4.16 node validates, relays
and mines exactly like 12.4.15.

P2P: pool links no longer stall at connect

  • Two pool nodes that connect push their pool catch-up at each other (jobs
    of 1.5-2.2 MB, shares, aux shares). Once the kernel buffers were full,
    both had writes pending, and a socket with a pending write was only
    watched for writing ("drain the write buffer first"): neither node read
    the other, and the link stayed dead (lastrecv = connect time, no pong,
    header or block) until the ping timeout. Seen on mainnet 12.4.14
    (120-123, 119-123, 123-124) and on testnet 12.4.15 (all links of a
    restarted node). A socket with pending writes is now also read while its
    receive queue has room (-maxreceivebuffer still bounds it), as Bitcoin
    Core does since v26. A TLS handshake step that must write first is
    unchanged.
  • While a peer's send buffer is full or getdata answers are pending, its
    pool messages are still taken, in order, and handed to the pool message
    thread; its other messages wait as before. Pool answers only ever go to
    -trustedblocksource peers, within their rate limits.
  • -testsocketbuffer=<bytes> (regtest, debug): kernel buffer size of every
    P2P socket. Test: p2p_pool_connect_burst_stall.py.

Pool

  • A block share is checked against the bits the chain requires at its own
    header time, not at the time it is verified. On testnet (minimum
    difficulty allowed 2 x the target spacing after the parent) a block share
    relayed, re-sent or retried more than 4 minutes after its parent was
    refused as pow and charged as a violation. Mainnet computes the same
    bits as before.

  • Honest pool peers are no longer disconnected for "invalid shares". The
    shares a node asks of every pool peer (missing shares of a pending or
    degraded block) were checked as the sender's own fresh work, so an answer
    carrying another node's shares, or our own old ones, scored a violation per
    share, and ten in a minute disconnected the peer (on mainnet about every
    2 minutes on some links; on testnet it triggered a 76-block split).
    Origin mismatches, our own shares sent back, late relayed shares and
    answers over the credit to ids we asked for are now refused without a
    violation. The same shares are accepted and refused as in 12.4.15 (no
    ledger or digest change); bad work, payout binding, malformed messages
    and floods are still violations. New counters: getpoolinfo
    peers[].rejects, peers[].violationsTotal, counters.rejectReasons.
    Test: p2p_pool_honest_rejects_not_punished.py.

  • No more blocks lost across a restart or a shutdown: NNG pool bridge jobs
    and their templates are kept in pool.sqlite (bridge_jobs) and restored
    at start, so a block share on a job issued before a restart is still
    accepted; the stratum submit thread and the ASIC Lotus and parent submit
    threads drain their queues at shutdown (10 s at most); a block-solving
    share, its job and its proof are on disk before the block is submitted;
    a worker's vardiff difficulty is restored after a reconnect or a restart
    (mining.sqlite worker_difficulty). New tables only, no schema version
    bump: 12.4.15 opens the files unchanged. Tests:
    feature_pool_bridge_restart.py, mining_stratum_shutdown_block.py,
    feature_pool_kill9_block_shares.py, mining_stratum_vardiff_restore.py.

Asynchronous explorer and module indexer

  • Writing the explorer database and the chain modules was about 77 % of
    "Connect block" on a 9,700-transaction block (408-424 of 533-550 ms), on
    the validation thread under cs_main. With -asyncindex (default 1)
    block connection and disconnection queue the block, and one worker
    applies exactly the same calls in the same order, off cs_main. The pool
    and mining modules (next template, stratum) stay on the validation
    thread. doc/async-indexer.md.
  • Regtest, same binary: "Connect block" median 444 ms with
    -asyncindex=0, 207 ms with the indexer; a fresh node syncing a
    50,000-transaction chain is fully indexed 1.7-2.3x sooner.
  • Crash safety without an on-disk queue: each index resumes from its own
    committed tip. Backpressure: at most -asyncindexmaxlag blocks (64) and
    -asyncindexmaxmem MiB (256) wait.
  • Reads: indexer {tip_height, index_height, behind, ...} in
    GET /api/v1/chain, /chain/sync, /health and the new
    getindexerinfo; REST waits at most -asyncindexreadwait ms (250) for
    the blocks queued before the request; a transaction or block still queued
    answers 503 explorer_indexing; the dashboard shows "indexing N blocks
    behind". syncwithvalidationinterfacequeue waits for the indexer;
    syncwithindexer waits for it alone.
  • -asyncindex=0 keeps the 12.4.15 path, with the same database contents.

Block connection and storage

  • Block and undo data are serialized once in memory and written with one
    call (after Bitcoin Core #31551); undo data is read in one call and
    checked in memory, falling back to the old path on anything unexpected.
    Same bytes on disk: a 12.4.15 node reads the files unchanged.
  • The UTXO set change of a large block is hashed in parallel slices
    (MuHash3072 products multiplied together). The result is bit-identical
    to 12.4.15 (multiplication modulo the MuHash prime does not depend on
    order or grouping; proof in doc/FORK.md), so the Napoleon UTXO
    commitments are unchanged. A 6,000-transaction block: UTXO set hash
    77.5 -> 21.4 ms, flush 86 -> 31 ms, connect block 158 -> 78 ms (regtest,
    loaded laptop, about +/-25 %).
  • Removing a connected block's transactions from the mempool is done in
    one pass (6,000 of 10,000 transactions: 31 -> 8 ms).
  • UTXO snapshots download in parallel chunks from several servers (2 M
    coins from three servers: 1.2 s instead of about 5 s). doc/utxo-snapshot.md.
  • SHA-256 uses the ARMv8 SHA2 instructions where the CPU has them (port of
    Bitcoin Core #24115), checked against the generic code at start-up.

Thin share relay

  • -thinrelay on a thin node serves LotusHash stratum miners with jobs
    from an upstream full pool node and forwards their shares over the NNG
    pool bridge; the full nodes keep every pool decision and verify every
    share (a relayed share gets exactly the weight of a direct one, and its
    proof of work must meet that weight). Relays are shown apart in
    getpoolinfo.relays, /api/v1/pool/relays, /api/v1/pool-public/relays
    and the fleet view, and never count as pool members. Upstream full nodes
    need 12.4.16 for the relay's exact share weight and views; older nodes
    ignore the new NNG fields. doc/thin-share-relay.md.

Batch transaction submission

  • sendrawtransactions rawtxs ( maxfeerate atomic ): sendrawtransaction
    for up to 1,000 transactions (8 MB) in one call. The batch is decoded,
    hashed and checked (context-free) on several threads, its scripts are
    verified in parallel with batch Schnorr outside cs_main (the
    -parallelmempool machinery), then the transactions are admitted in ONE
    cs_main / mempool pass in the order given (a parent before its
    children), with one notification burst, one mempool trim and one
    periodic flush. The fee cap is checked in that same pass (no second
    test_accept pass as in sendrawtransaction).
    • Best effort (default): one result per transaction (accepted,
      already_in_mempool, fee, or the code / error /
      reject_reason sendrawtransaction would give).
    • atomic=true: all or nothing. If one transaction is refused, or the
      batch would push the mempool over -maxmempool, the ones already added
      are removed again without any notification and without taking mempool
      sequence numbers: listeners (wallet, ZMQ, explorer, token overlays) and
      getrawmempool's mempool_sequence never see them.
    • The same txid twice in a batch is refused (duplicate-in-batch).
  • REST POST /api/v1/txs/broadcast-batch ({"txs": [...], "atomic": false, "maxfeerate": n}), rate limited node-wide: -restbatchbroadcastrate
    batches (default 30) and -restbatchbroadcasttxs transactions (default
    10,000) per minute, 429 beyond, 403 when the rate is 0. Not in the NNG REST
    tunnel, not in the unauthenticated /api/v1/rpc allowlist, no token burn
    guard. Keep it closed on public proxies unless opened on purpose
    (doc/production-deployment.md, "Batch broadcast").

Mining templates

  • getblocktemplatelight / submitblocklight "hexdata" "job_id" (the
    BCHN getblocktemplatelight idea): the template without its transactions,
    with job_id, the coinbase merkle branch (merkle), txcount and
    sizewithoutcoinbase; the pool submits the header and its own coinbase
    only, and the node appends the job's transactions. The same template
    always gets the same job id (two nodes with the same mempool agree).
    Jobs are kept for the last -gbtlightjobs templates (default 64).
  • getblocktemplate and getblocktemplatelight answer from a published
    snapshot, without cs_main, while the tip and the mempool update counter
    are those of the template (the 5 s rebuild rule of the shared template is
    unchanged). On min-difficulty chains (testnet, regtest) the snapshot
    serves the second it was taken in, since nBits follows the header time.
  • Opt-in in 12.4.16, the defaults in a later release. The new block
    selector and the larger chain limits below are available as options;
    12.4.16 keeps the 12.4.15 defaults (-blockselector=ancestor, 50 / 101
    kB). They become the defaults in a later release, once every node runs
    12.4.16: peers on 12.4.15 or older refuse transactions past depth 50, so
    deeper chains only propagate between 12.4.16 nodes, and a mix of
    selectors across the fleet would change which transactions get mined.
  • Block selector (-blockselector, policy only): feerate (opt-in;
    the ABC D13116 / BCHN v23 idea) picks each transaction on its own
    modified fee rate, best first, a child only once all its in-mempool
    parents are in, ties by txid; O(N log N). No CPFP: a high-fee child does
    not pull in a low-fee parent (below -blockmintxfee, or when the block
    is full, both stay out). ancestor (default) is the 12.4.15 CPFP
    package selector. The block's transaction order is the canonical one
    either way. Pool payouts are unchanged in form with either: the coinbase
    value is the subsidy plus half the fees of the selected transactions,
    and the selection is a function of the mempool (stratum, NNG and the
    pool hook all use it).
  • Mined chains removed in linear time (policy-neutral): removing a
    connected block's chain from the mempool no longer walks every removed
    entry's descendants (a 2,000-transaction chain mined whole: 1,050 ms ->
    51 ms; 500: 62 -> 9 ms). This is what makes larger chain limits workable.
  • Larger chain limits, opt-in: -limitancestorcount=500 -limitdescendantcount=500 -limitancestorsize=500 -limitdescendantsize=500 (defaults unchanged: 50 / 101 kB). Admitting a
    transaction still costs O(its in-mempool ancestors): 0.3 ms at depth
    500. -maxmempool must hold 40 packages of -limitdescendantsize, as in
    12.4.15: 5 MB with the default (unchanged), 20 MB with 500 kB.

Measured (batch submission and templates)

Regtest, laptop shared with other jobs, same binary and same run per pair:
sendrawtransactions admits 2.2-2.7x more transactions per second than
JSON-RPC batch arrays (A1, 6,000 signed P2PKH, b=200: 13.8-16.6k -> 33.9-39.5k
tx/s). With a 6,000-transaction mempool, a cached getblocktemplatelight
answers in 0.38 ms (p50) against 28-36 ms for the full template, and 12
readers get 12,700 req/s against 167-313. Block assembly with feerate:
130-134 ms against 136-143 ms with ancestor (TestBlockValidity
dominates), with the mempool lock held about half as long. Full numbers:
contrib/bench (bench_ingest.py, bench_reads.py) and the 12.4.16 benchmark
report.

Node web UI: mining views by scope, unified docs

  • Mining views by scope: the Pool tab has Pool (fleet), This node
    (/dashboard/pool/node; the old /dashboard/mining URLs lead there) and
    Network. Merged hashrate (LotusHash + SHA-256d/M + Scrypt/M) for the node,
    the pool and the network, with a LotusHash badge on native figures;
    pool-wide figures per algorithm; each figure has one definition
    (doc/mining-figures.md). The fleet tile shows thin nodes and thin share
    relays apart: they are never counted as pool members.
  • REST: pool-public/summary.algorithms, hashrate_merged_lotus_equivalent
    (mining/status, pool-public/summary, mining/network), mining/status.mmv2,
    mining/aux/workers counts.by_algo_state, algo on pool-public/blocks
    rows, /api/v1/mining/hashrate/series per algorithm.
  • pool.sqlite and mining.sqlite record the merged-mining algorithm of every
    share, block and series point (columns added with a default; no schema
    version bump: a 12.4.15 binary still opens the files; the ledger digest
    is unchanged). doc/mining-sqlite.md.
  • Documentation: the node's docs are Markdown pages embedded in the binary
    and rendered in the dashboard's Docs tab (sections, "since" markers for
    12.4.16 features, links checked at build time); the lotusia.network site
    shows a pinned copy of the same pages. doc/documentation.md.
  • A direct link to a block or developers page now gets every card: the
    first route waits for all of the page's scripts (the UTXO commitment card
    was missing on a cold load).

Taproot (P2TR) addresses

A Taproot address (lotus_J…, 57–58 characters) can now be created,
received on and spent from everywhere, not only watched. doc/p2tr.md has
the rules, the derivation paths and the support table.

  • Node wallet: getnewaddress "" taproot (alias p2tr),
    getrawchangeaddress taproot. The wallet owns the key-path-only Taproot
    address of each of its keys and spends it with one Schnorr signature
    (sendtoaddress, signrawtransactionwithwallet,
    signrawtransactionwithkey, PSBT), and re-signs a pending Taproot spend
    at First Chronicles like any other. getaddressinfo shows the commitment, the
    internal key and a tr() descriptor; dumpprivkey of a P2TR address
    returns its internal key. New tr(KEY) descriptor (scantxoutset,
    deriveaddresses); descriptor wallets get tr(xpub/86'/…), so
    getwalletinfo reports twice the key pool size for them. The default
    address type stays legacy and change stays on legacy addresses.
  • Mining and pool: a P2TR address works as generatetoaddress /
    getrawunsolvedblock address, -stratumcoinbase, stratum user name and
    ASIC payout address, and is paid by the pool (36-byte script, already in
    the payout whitelist). The internal taproot:<hex> spelling of a user
    name is stored as the lotus_J… address (one account per miner). The
    public template service accepts a P2TR payout script.
  • Stricter: taproot:<hex> whose 33 bytes are not a valid public key
    is refused everywhere (it used to pass and could burn a reward); the
    lotus_J… form already was. -ltccoinbaseaddr / -dogecoinbaseaddr
    ignore a Lotus Taproot address.
  • Gate: wherever a P2TR address is handed out, accepted as a payout or
    paid by the wallet, it is refused with the reason while the next block
    does not accept Taproot (regtest and new test networks before
    -secondsamuelactivationtime, a node in initial sync). Always open on
    mainnet. GET /api/v1/chain has a new taproot {active, activation_time} object.
  • REST / explorer: mempool transaction outputs report their real
    script_type (was pubkeyhash for every addressed output);
    /api/v1/addresses/taproot:<hex> is looked up under the lotus_J…
    address.
  • Web wallet: Accounts › Address type: Legacy (P2PKH) or Taproot
    (P2TR), each with its address, path and balance; receive, send to and
    spend from Taproot in the browser. Taproot derives at
    m/86'/10605'/0'/0/i from the same recovery phrase. Fixes addresses on
    the Napoleon test network (lotusT…, was lotus_…).
  • Dashboard: a plain payment to or from a P2TR address is no longer
    labelled a contract (no "Contract created", no key-path warning); the
    address page shows the address type; search understands taproot:<hex>
    and a bare commitment.
  • GPU miner CLI 0.4.x and Lotus Miner app: the payout address (and a
    Lotus address used as a stratum user) is checked before mining: Legacy,
    Script or Taproot, typos, cut copies, another network, a Taproot address
    without a valid key. The CLI exits with code 2 and the reason; a node or
    pool that refuses Taproot gets a plain explanation. The address checks
    live in one crate, gpuminer/lotus-address, used by both.
  • xpi-ts 0.3.2-p2tr.1 (vendored in the wallet): P2TR helper; a
    Taproot key-path input is now only signed by its own key.

Decided for this release: the web wallet derives Taproot addresses at
m/86'/10605'/0'/0/i (final), and the default address type stays Legacy
(P2PKH) in the web wallet and in the node wallet; Taproot is opt-in.

GPU miner and Lotus Miner app

  • One GPU miner for NVIDIA and AMD: OpenCL is loaded at run time, so
    one binary (lotus-gpu-miner-<v>-linux-x86_64.tar.gz) and one image
    (ghcr.io/lotusdevorg/lotus-gpu-miner, boblepointu/lotusgpuminer)
    serve both vendors. --backend auto|nvidia|amd (default auto: every
    OpenCL device, numbered as in earlier releases, so -g values, rigs and
    app settings keep their meaning), --list-devices; the app has a GPU
    backend selector. The lotus-gpu-miner-nvidia-… / lotus-gpu-miner-amd-…
    archives, the lotus-gpu-miner-amd image and the :<tag>-amd Docker Hub
    tags are no longer published; their old versions stay. The image is
    575 MB unpacked (268 MB for the NVIDIA-only image). NVIDIA hashrates are
    unchanged (RTX 4080 Laptop, LotusHash, SHA-256d, Scrypt). AMD is
    untested on real hardware
    in this release (no AMD GPU was available):
    device selection is unit-tested and the image carries Mesa OpenCL.
  • No built-in payout address: the miner, the image and the app never
    mine to a default address any more. Without an address the CLI and the
    image launcher exit with code 2 and say how to pass one (-o,
    MINER_ADDRESS); a config.toml still equal to the old template is
    replaced. Without a node, LotusHash goes to the FRENCHBTC pool (the
    fleet's stratum ports, in failover order). The dead burnlotus.org
    defaults are gone.

Upgrade notes

Nothing to do for a node. A descriptor wallet created by an older version
has no tr() descriptor; create a new wallet or import one. GPU miner
users who relied on the built-in address must now pass their own
(-o <address> or MINER_ADDRESS); the vast.ai and home rig command lines
already do.

Operators of a public proxy: keep POST /api/v1/txs/broadcast-batch
closed unless it is opened on purpose (doc/production-deployment.md,
"Batch broadcast").

SHA-256 of every file
5f61617bbd3e09218822140e47b64e67be3bd6d3bfd69211096043440e0a0dcc  lotus-binaries-12.4.16-linux-x86_64.tar.gz
91b6a51711f17fd2ef0e19cfc863d4884d101eed88b5458e511815259f256dc6  lotus-binaries-12.4.16-macos-x86_64.tar.gz
d5d504f4d4f2f7a6ffbf530dcc34f7850a16a2a0fe0d4ee4e36b7a271e8acc1a  lotus-cli-12.4.16-linux-x86_64.tar.gz
d7039e8726d380d349a706838f3959161c21133fa696fcedadcc1ec89af1f6e9  lotus-cli-12.4.16-macos-x86_64.tar.gz
0d9adbf254f51ab479e1c0121f8ed744cefbdbba981cdd75c3f54a0d33a5bacc  lotus-gpu-miner-12.4.16-linux-x86_64.tar.gz
4c22e3c3ef0ea0ac3d15909295642cc0185fa7f9c31003b1cad57672feed4e12  lotus-miner-gui-12.4.16-linux-x86_64.tar.gz
dcf895285e8356887422cc19080bce48f2ac987bfa60cbffb0a36637da80b7ba  lotus-seeder-12.4.16-linux-x86_64.tar.gz
a319c4e74dd98621e5658bc9dc052fdee3073c1bc74e69e21162bab9227a30be  lotus-seeder-12.4.16-macos-x86_64.tar.gz
1766debdc3354e975b99b90a5f1da49a84bc1ea839d7dd3b6ab731b4bf4ac5fe  lotus-tx-12.4.16-linux-x86_64.tar.gz
e85bfb7dbe6eef506630867e1053dfb34cc4d550763d4bd2177369680a719a15  lotus-tx-12.4.16-macos-x86_64.tar.gz
0c49e1162cb6859c856f96d8eff14564c99f8a1b1a56f46e60ff002ce0a837bc  lotus-wallet-12.4.16-linux-x86_64.tar.gz
ab366adde6238bbf7c1521a95e3c924b5e648621696dc1980330fe77584cba9c  lotus-wallet-12.4.16-macos-x86_64.tar.gz
0f9b7c8aa2e0866094ed576b7ca04e7c3ba46597d31c6fd87a153dbdd14da623  lotusd-12.4.16-linux-x86_64.tar.gz
ed9926b7919bb4b0cad85ab377328b3fdd006073f8b0130cf1a949e31093329d  lotusd-12.4.16-macos-x86_64.tar.gz
a29583c1458fc230179576ef59d152d3f1596d9449aad0594716ece2aa7f0674  lotus-binaries-12.4.16-windows-x86_64.zip
730570007e031d9e5375e64ef7273adfe3818d1db61fdea8e87c91370b81a25c  lotus-cli-12.4.16-windows-x86_64.zip
2613cb7804b33dae750b04b856d3aecd5b6b5e9719a897cb7c5b9ed6ebf351ca  lotus-miner-gui-12.4.16-macos-arm64.zip
a0a82d785e663f1f40928d4b02894acf61396f7f0432e956625b853fb2d77375  lotus-miner-gui-12.4.16-macos-universal.zip
4e82eea742fbdba88b681863f668d43b21b0beb994de61856072454ae8cdc662  lotus-miner-gui-12.4.16-macos-x86_64.zip
a9811807fc3f399c0ec6219088cfa59b7b31deb7f76d74cafa2c40511aed6730  lotus-miner-gui-12.4.16-windows-x86_64.zip
40b83f2a821f6f58bcdd866be392b02970239df9f212c541f8ac953526a28d0b  lotus-tx-12.4.16-windows-x86_64.zip
3d4afae80a7e0932e331aa4fd0f8b175b8619101abfaf30e09f43242a8d321f6  lotus-wallet-12.4.16-windows-x86_64.zip
33befd3eb474accef36bcd737fe78762fb06f84a2616dce31470db97f31b9cc7  lotusd-12.4.16-windows-x86_64.zip

Consensus changes: none. The rules and the activation time are those of 12.4.13 · Minimum version on the network: 12.4.13, required from Napoleon (2026-12-21 20:50 UTC)

Install and upgrade guide · Web wallet · All releases · Container images · Report a problem