Bitcoin Core v31.0 shipped on April 19, 2026 — the most architecturally significant upgrade to the reference client since SegWit. The release replaces the decade-old ancestor/descendant mempool topology with a cluster-based system that caps transaction groupings at 64 transactions or 101 kB, rest...
"Thanks everyone for getting us this far." — Pieter Wuille (sipa), Bitcoin Core Developer, on the completion of the cluster mempool working group
Bitcoin Core v31.0 shipped on April 19, 2026 — the most architecturally significant upgrade to the reference client since SegWit. The release replaces the decade-old ancestor/descendant mempool topology with a cluster-based system that caps transaction groupings at 64 transactions or 101 kB, restructures replace-by-fee (RBF) validation around feerate diagrams, and introduces native Tor/I2P-only transaction broadcasting via a single configuration flag. Over 175 developers contributed to the release.
The upgrade addresses two distinct failure modes in Bitcoin's base layer: inefficient block construction that has cost miners measurable revenue, and a privacy gap in transaction relay that chain-analysis firms have exploited since Bitcoin's early years. Both fixes arrive as the network supports 24,557 reachable nodes and processes roughly 328,372 BTC held by the U.S. government alone — infrastructure that institutional participants increasingly depend on.
For Layer 2 protocols, particularly the Lightning Network, cluster mempool eliminates a class of transaction-pinning attacks that have been documented since 2020 and never fully resolved under the old architecture.
Since Bitcoin's earliest versions, the mempool — the waiting room for unconfirmed transactions — enforced a maximum of 25 ancestors or 25 descendants per transaction. This topology-based approach made transaction selection for block construction an approximation at best. Miners building blocks through getblocktemplate RPC calls received templates that were not optimally ordered by fee, because the node could not efficiently evaluate packages of related transactions as a unit.
Bitcoin Core v31 replaces this system entirely. Transactions in the mempool are now organized into "clusters" — connected components of parent-child relationships — bounded at 64 transactions or 101 kilobytes of virtual size. Within each cluster, transactions are sorted into "chunks" ordered by effective feerate. Two new RPCs expose this structure: getmempoolcluster returns the cluster a given transaction belongs to, while getmempoolfeeratediagram provides a full view of the mempool's fee distribution.
The concept traces back to a 2023 proposal by Bitcoin Core developers Pieter Wuille and Suhas Daftuar. The code was integrated into the Bitcoin Core main branch on November 25, 2025, via PR #336292, following over three years of design, review, and testing. Developer "instagibbs" characterized the effort as a "well fought project."
The practical consequence: block templates generated by v31 nodes should capture more fee revenue than templates from v30 nodes, because transaction ordering now reflects actual mining incentives rather than topology heuristics. For miners operating at thin margins — and with hashrate increasingly diverted to AI compute workloads, as covered in prior reporting — this is a material efficiency gain.
The old RBF rules evaluated transaction replacements on a per-transaction basis: a replacement was accepted if it paid a higher absolute fee than the transaction it replaced, plus enough to cover the relay cost. This created edge cases where accepting a replacement actually made the mempool worse — a higher-fee replacement could displace a package of transactions whose aggregate feerate was superior.
V31 replaces this logic with a feerate-diagram test. A transaction replacement is now accepted only if the resulting mempool's feerate diagram is "strictly better" than the diagram before the replacement. According to the release notes, this "eliminates all known cases of replacements occurring that make the mempool worse off."
The distinction matters for MEV dynamics on Bitcoin. Under the old system, sophisticated actors could construct replacement transactions that degraded mempool quality while extracting value — a form of griefing that was difficult to prevent. The feerate-diagram approach aligns replacement incentives with miner incentives: if a replacement does not produce a more profitable block, nodes reject it.
Additionally, the CPFP (Child Pays for Parent) carveout — a special exemption that allowed one extra child transaction beyond normal limits — has been removed. Smart contracts and protocols that relied on this behavior are directed to use TRUC (Topologically Restricted Until Confirmation) transactions instead, a more structured approach to package relay.
The minimum fee rate bucket for estimation has also been lowered from 1 sat/vB to 0.1 sat/vB, matching the default minrelaytxfee. Fee estimation accuracy has been a persistent pain point: a March 2025 analysis documented variations of 35–127% between estimation tools during identical time periods. The lower bucket floor and cluster-aware estimation should narrow these spreads.
Transaction pinning has been a known vulnerability in Lightning Network penalty mechanisms since at least 2020. The attack works as follows: a malicious counterparty broadcasts a low-fee transaction that is related to a channel closure, then "pins" it in the mempool by making it part of a transaction package that is expensive to replace under the old RBF rules. The honest party's penalty transaction cannot propagate to miners, potentially allowing the attacker to steal funds after the timelock expires.
Cluster mempool addresses this by unifying eviction, inclusion, and replacement logic around a single feerate-ordering principle. Because the mempool now evaluates transactions within clusters rather than individually, a pinning transaction that degrades the cluster's aggregate feerate will be evicted or replaced more reliably.
The removal of CPFP carveout and the introduction of TRUC transactions provides Lightning implementations with a more predictable mechanism for fee-bumping commitment transactions. Under the old system, CPFP carveout was the primary tool for ensuring that channel closure transactions could be accelerated — but its exemption from normal limits made it fragile and difficult to reason about in adversarial scenarios.
These changes do not eliminate all forms of transaction pinning. But they reduce the attack surface meaningfully for Lightning and other Layer 2 protocols that depend on timely transaction confirmation for security. Lightning developers have been tracking the cluster mempool project for years; the production release removes a longstanding caveat from Lightning's security model.
Bitcoin Core v31 introduces the -privatebroadcast configuration option. When enabled, sendrawtransaction RPC calls are routed exclusively through Tor or I2P networks. The node's clearnet IP address never touches the transaction relay path.
This addresses a specific surveillance technique. Chain-analysis firms operate thousands of listening nodes across North America and Europe, according to industry estimates. These nodes record the IP address of the first peer to relay a given transaction, linking network identity to on-chain activity through timing analysis. The technique has been used in law enforcement investigations and commercial compliance products since at least 2014.
Previous versions of Bitcoin Core could be configured to route traffic through Tor, but the process required running a separate Tor proxy, manually configuring the node, and ensuring no clearnet connection leaked during broadcast. V31 reduces this to a single boolean flag in bitcoin.conf. Supporting RPCs — getprivatebroadcastinfo and abortprivatebroadcast — provide operational visibility into the feature's status.
The privacy floor for average node operators rises significantly. A user running a default v31 node with -privatebroadcast=1 achieves network-level privacy that previously required technical expertise. Whether this changes the calculus for regulators — particularly in jurisdictions implementing the Crypto-Asset Reporting Framework (CARF) — remains an open question.
It is worth noting that this feature protects the link between IP addresses and transactions, not on-chain transaction tracing. Address clustering, output analysis, and other on-chain heuristics are unaffected. The feature is complementary to, not a replacement for, CoinJoin and other on-chain privacy techniques.
Several infrastructure-level changes accompany the mempool and privacy upgrades:
Memory allocation. The default -dbcache has been increased from 450 MiB to 1,024 MiB on systems with 4 GB or more of RAM. This accelerates initial block download (IBD) and ongoing block validation. Node operators on resource-constrained hardware retain the old default.
GUI framework. The graphical interface has been updated from Qt 5 to Qt 6.8. The console history now filters sensitive wallet commands, reducing the risk of credential exposure through log files.
REST API. A new /rest/blockpart/ endpoint enables byte-range fetching of blocks, allowing lightweight clients to request specific portions of a block without downloading the entire data structure.
Fee estimation floor. The minimum feerate bucket has been lowered from 1.0 sat/vB to 0.1 sat/vB, providing more granular estimation during low-fee periods.
Compiler requirements. Minimum supported compilers have been raised to Clang 17.0 and GCC 12.1, reflecting the codebase's adoption of modern C++ features.
Deprecated features. The paytxfee and maxorphantx configuration options have been removed. The old Tor network label has been deprecated in favor of updated naming conventions.
Bitcoin Core v31.0 was released on April 19, 2026, four days ago as of this writing. The rc4 release candidate entered testnet on April 11. Historically, Bitcoin Core adoption follows a gradual curve. When v26 launched, approximately 1% of nodes upgraded on launch day. Adoption of v30 was projected to reach 40–50% by Q1 2026.
The network currently runs approximately 24,557 reachable nodes according to Bitnodes data, distributed across the United States (2,695), Germany (1,241), France (678), Finland (404), Canada (376), and other jurisdictions. Not all of these run Bitcoin Core — approximately 15–20% use Bitcoin Knots or other implementations — but Core remains the dominant client.
For the cluster mempool changes to affect network-wide behavior, a significant percentage of nodes must upgrade. Transaction relay policies are enforced locally by each node; a v31 node will evaluate and relay transactions according to the new cluster logic regardless of what its peers run. However, the full benefits — particularly for Lightning pinning resistance — accrue only when most mining nodes run v31.
Mining pools, which produce the majority of Bitcoin blocks, tend to upgrade faster than individual node operators due to the direct revenue implications. The improved block template construction provides a straightforward economic incentive to upgrade.
Bitcoin Core v31.0 is an infrastructure release, not a feature release in the user-facing sense. There are no new opcodes, no changes to the consensus rules, no soft forks. What changed is the machinery underneath: how nodes organize transactions, how they decide which replacements to accept, and how they protect the privacy of transaction originators.
The economic implications are concentrated among two groups. Miners receive better block templates — a direct revenue improvement at a time when Bitcoin's security budget faces pressure from hashrate diversion to AI workloads. Lightning Network operators gain a more predictable base layer, with fewer edge cases in which penalty transactions can be pinned by adversaries.
For the broader network, the privacy upgrade is the most broadly applicable change. A single flag in a configuration file now provides IP-level privacy that previously required significant technical setup. Whether this accelerates or complicates the regulatory trajectory for self-hosted nodes will depend on jurisdiction-specific interpretations.
The release represents over three years of development. Its impact will unfold over months as nodes upgrade. The data to watch: mining pool adoption rates, Lightning channel closure success rates, and fee estimation variance across the network.