Algorand's v5.0.0 clears 90% node support; native Falcon-1024 accounts head to mainnet

2026-08-17

protocol-upgrade · quantum-resistance · fees · developers · mainnet · discovery · web · updated

Algorand's v5.0.0 clears 90% node support; native Falcon-1024 accounts head to mainnet

A Supermajority in Three Days

Three days after Go-Algorand 5.0.0 was published, the network's node operators had already done what usually takes the community weeks: push the new software past the threshold that converts a release into Algorand's live consensus rules. It is the largest protocol upgrade since staking rewards landed in January 2025, and it delivers the first milestone of the post-quantum roadmap the Algorand Foundation published in June. In the Foundation's confirmation: 'The v5.0.0 upgrade reached the needed 90% support threshold on August 15. It will go live on [mainnet](/glossary/mainnet "The main public blockchain network where transactions are recorded and have real value, as opposed to a test network used for experimentation.") after a mandatory 208,000 round (~7 day) cooldown period.' That puts [on-chain](/glossary/on-chain "Describes transactions that are recorded and permanently stored on a blockchain ledger, making them publicly visible and tamper-resistant.") activation around August 22 — and the threshold is measured by node adoption, not a Foundation ballot, so the vote of approval was effectively cast by the operators running the network.

The change is one story told in three parts: quantum-resilient accounts become native, transaction fees start reflecting what a transaction actually contains, and smart contracts get a materially bigger ceiling.

Post-Quantum Accounts, Native at Last

The headline feature answers a threat most of the industry is still planning for. Almost all of today's crypto security rests on [elliptic](/glossary/elliptic "A blockchain analytics and compliance firm whose screening is integrated into the Aid Trust Portal, so wallets that have interacted with suspicious ac")-curve math that no existing computer can break, but a sufficiently powerful quantum computer running Shor's algorithm could. Researchers call the long game 'harvest now, decrypt later' — hostile actors recording signed or encrypted data today in order to crack it decades from now — which is why hardening the ledger's past matters for assets held in the present. Falcon, the scheme Algorand chose, is a lattice-based post-quantum signature standard selected by NIST; the chain has used it since 2022 in its [State Proofs](/glossary/state-proofs "A compact, quantum-resistant cryptographic proof that certifies the state of a blockchain ledger at a certain round. Algorand uses Falcon-based state "), compact quantum-resistant snapshots produced every 256 rounds that anchor the ledger's entire history back to genesis, and it carried its first post-quantum mainnet transaction in 2025.

Until this release, a quantum-resistant account had to be assembled from a LogicSig — a stateless program the [AVM](/glossary/avm "Algorand Virtual Machine, the software environment that executes smart contracts on the Algorand blockchain.") (the Algorand Virtual Machine, where smart contracts execute) runs that bakes Falcon verification into the account logic. It worked, but in the Foundation's own framing those accounts were 'not natively supported by the ledger, developer tooling, or consensus rules.' Version 5.0.0 makes [Falcon-1024](/glossary/falcon-1024 "A digital signature algorithm designed to be secure even against future quantum computers, standardised by NIST at its highest security level.") a first-class way to authorize an account, on the same footing as the Ed25519 signatures every address has used since genesis. The structural problem was key size: a Falcon-1024 public key is 1,793 bytes against a 32-byte address. Algorand's solution derives the address as a hash of a protocol domain separator, a two-byte scheme identifier, a one-byte salt and the public key, preserving the familiar 32-byte, 58-character address format without bloating the ledger. The salt is chosen by rejection sampling so that 'a post-quantum address is one that no Ed25519 private key could ever control' — even a forger who breaks elliptic-curve math has nothing to sign against it.

Tooling moves in the same window. The release adds algokey pq commands for key management, and SDKs, Pera and AlgoKit are expected to support deriving Falcon-1024 accounts from the standard 25-word mnemonic. Two costs deserve honesty. Falcon signatures cannot be batch-verified the way Ed25519 can, so each one is verified individually, and they occupy far more block space — which is precisely what the new fee model charges for (more below). What stays classical, for now, is the consensus machinery itself: block proposal and voting still run on Ed25519 keys, a post-quantum VRF research paper is not due until early 2027, and hardware wallets are not Falcon-capable at this release. The Foundation does not claim otherwise: it states plainly that no system is ever completely future-proof and that other parts of the protocol are still migrating.

Fees That Reflect What a Transaction Carries

This is the first Algorand upgrade where the fee you owe depends on what your transaction actually contains. The network's effort is now expressed as 'usage,' a fixed-point multiplier on the minimum fee measured in millionths: an ordinary transaction has a usage of 1,000,000 and pays one minimum fee, while a Falcon-authorized transaction adds 2,000,000 — three minimum fees in total — because its signature and public key consume so much extra block space. The engineers who built it led with a compatibility promise: 'No transaction that is legal today will cost more under these new rules.' Every free allowance is at least what it was before, and everything above the allowance is surcharged at one ten-thousandth of a minimum fee per byte, about a tenth of a microAlgo per byte at today's minimum fee.

Concept	Real-World Implication
Note field: 1,024-byte free allowance, 4,096-byte new limit	A maximum-length note rises from 1,000 to 1,308 microAlgos — the largest surcharge, at +0.31 minimum fee
Application arguments: 2,048-byte free allowance, 16,384-byte new limit	Complex app calls can carry far more data for at most +1.43 minimum fees
Application programs: 8,192-byte free allowance	Extra program pages beyond the old cap are now buyable, max surcharge +0.82 minimum fee
LogicSig program: 1,000 bytes per transaction, pooled across a group, now 16,000 bytes	A 16-KB program costs 2,434 microAlgos (about 2.5 minimum fees) in a single transaction instead of padding a group with fifteen dummy transactions

Fees still pool across a group and round up exactly once, and the tooling story matters as much as the pricing: dApps that previously multiplied the minimum fee by their transaction count must now read group usage from the simulate endpoint. The guidance is explicit — 'Do not hardcode fees. Read min-fee from /v2/transactions/params, and read group-usage from simulate.' There is also a deliberate preview of the future in this release: block headers now carry a Load value (how full the block is, in millionths) that feeds a CongestionTax modeled on EIP-1559, targeting half-full blocks — but nothing charges that tax yet. It is measured, validated, and ignored, so the ecosystem can watch it move during real traffic before it ever affects fees.

The fee model is also the Foundation's first concrete answer to a debate that has been running on the Algorand forum all year under the banner of the upcoming 'King Safety' paper: how the network pays for itself once Foundation subsidies run down. Community sentiment has been skeptical — user Maugli summarized the mood in April: 'The technology is good. The economics are broken.' — and builder Ludo (scholtz) warned in August that a fee-centered fix could miss the underlying problem, since his reading of two tracked Foundation addresses puts roughly 154M ALGO sold into the market over the past year against fee revenue that even generous throughput assumptions cannot match. V5.0.0 deliberately sidesteps that fight: it does not raise fees on anything legal today, pricing only the new resource-heavy features, while quietly installing the machinery a future congestion charge would need.

A Bigger Toolkit for Builders

Developers get the largest AVM update in years, and it is financed by the same usage model. The smart-contract program ceiling doubles, from 8,192 to 16,384 bytes, via up to seven extra 2,048-byte pages — and, critically, the page count is no longer frozen at creation. An update transaction can now change an app's program size and global-state schema, letting a live application grow in place while keeping its ID, its address and its state instead of forcing a redeploy that would strand every box and asset holding. That raised a payment question the protocol answers with a new on-chain concept, the SizeSponsor: the minimum-balance obligation for extra pages moves to whoever actually pays for the growth, readable via app_params_get AppSizeSponsor, so a project can be maintained long after its original creator is gone.

[Box storage](/glossary/box-storage "On-chain key-value state attached to a smart contract, used by AlgoDirectory to store listing data directly on the blockchain instead of in a traditio"), the dedicated per-app data store, was previously private — invisible to every other application. Nine new opcodes (app_box_create, app_box_extract, app_box_replace, app_box_del, app_box_len, app_box_get, app_box_put, app_box_splice, app_box_resize) let one application reach into another's boxes, but only under two opt-in flags the owning app's own code must set: ForeignBoxReads for read-only access, and FamilyBoxAccess for read-write among apps sharing a creator. The restriction is not paranoia: an unrestricted write permission would let a hostile app fill a stranger's boxes until the entire account balance was locked, since box minimum balances are always charged to the owning app. Shared mutable state also reopens the re-entrancy hazard the AVM had designed away — a call chain that lets a sibling app write state mid-update — so the protocol re-closes it with a family-scope write ban rather than leaving it to developer discipline.

AVM v13 adds two hash functions for the cryptography crowd: full sha512 (all 64 bytes, verifiable against the new SHA-512 block commitments) and poseidon2, an arithmetization-friendly hash with BN254 or BLS12-381 parameter sets that is cheap to prove inside a zero-knowledge circuit — a step toward native ZK development. The long-deprecated dryrun endpoint and tealdbg tool are removed in favor of simulate, and a new GET /v2/node/peers endpoint exposes the peer graph to operators.

Node Runners: The Clock Is Ticking

The activation window is real, and the infrastructure release schedule reflects it. Algorand Indexer v3.10.0 and Conduit v1.10.0 were published August 10 specifically to support the new consensus, and runners must upgrade both installs to remain on the network once the cooldown ends. NodeKit users can run ./nodekit upgrade (from NodeKit v1.1.0 onward this also upgrades NodeKit itself), the FUNC Program updates automatically, and custom setups are directed to the Foundation's devrel address. The scale of the operator base is visible in Nodely's telemetry: roughly 2,635 full-time mainnet nodes as of August 16, slightly below the 2,683 measured on release day — a small dip consistent with a fleet in the middle of rolling over to new software.

The Roadmap Continues

V5.0.0 is the first deliverable off the June roadmap, not the destination. Still ahead by the Foundation's own schedule: Falcon-512, whose roughly half-size signatures suit shorter-lived keys, by year's end; post-quantum [multisig](/glossary/multisig "Short for multi-signature, a security setup where more than one private key is required to approve a transaction. A 3-of-5 multisig means any three of") by the end of 2026, with m-of-n quorums mixing classical, pure-Falcon and hybrid keys for treasury-grade custody; a post-quantum VRF research paper in early 2027; and then the hard part — migrating consensus itself. What ships today is the foundation those pieces stand on: a ledger that treats Falcon as a first-class citizen, a fee model that can pay for the space it consumes, and an ecosystem debate, finally, with concrete machinery to argue about.

Source

Source: https://forum.algorand.co/latest