Cardano merges key Leios code, but the 15x scaling is a target that still needs votes and pool keys

The Cardano developers who maintain the node’s consensus code merged the core of Ouroboros Leios, the throughput upgrade known as Cardano Leios, into their main branch on 9 October 2026. The change came in pull request #2386 on the IntersectMBO ouroboros-consensus repository, which introduces a protocol called Praos2, described in its title as “Praos plus the Leios overlay”. The consensus team’s weekly report of 23 October confirmed the merge and said two networking parts had followed it into the main branch.

The code is real progress. The headline number attached to it is not a result. CryptoSlate, which wrote up the merge on 26 October, put a “15x scaling breakthrough” in its headline, but the 15x figure is the top of a conditional range published by Input Output (IOG): Leios is meant to start at three to five times today’s capacity and reach roughly 10 to 15 times only after at least two later parameter updates, each of them a governance vote. No date exists for any of those steps, and one piece of the software, the selection of endorser blocks, is still unwritten.

Cardano (ADA) mainnet still runs plain Praos. Between the merged code and a faster chain stand a testnet that has not yet switched to the new rules, a constitution amendment in draft, a hard fork vote, key registration by several hundred stake pools and at least two more parameter votes.

Cardano Leios code: what was merged on 9 October

Pull request #2386 was opened by the GitHub user nfrisby and approved by the reviewers ch1bo and jasagredo before it was merged as commit 35d0289. ch1bo had first asked for changes. The pull request was written as an alternative to an earlier one, #2354, which ch1bo said in the thread he was closing. In practice, Praos2 is the existing Ouroboros Praos protocol, which produces Cardano’s blocks today, with an extra layer on top that lets the chain carry far more transaction data.

Several smaller merges followed on the same repository in the next two weeks:

• #2400, “Leios: integrate latest changes for 2026w41”, merged on 11 October.
• #2415, which stops Praos2 from certifying across epoch boundaries, merged on 16 October.
• #2278, which adapts the formal specification to Linear Leios, the current design, merged on 20 October.

A fourth merge, #2427 on 21 October, adds a ledger view for Peras, a separate faster-finality project. It is labelled Peras on GitHub, not Leios, and should not be counted as Leios progress.

The networking pieces reached main by 23 October

The weekly development report dated 23 October on Essential Cardano says Praos2 “had merged into main” and that the LeiosNotify clients and LeiosFetch servers had moved there too. Those are the two halves of how nodes tell each other that new Leios data exists and then download it. The report also says the prototype now runs stricter checks on what peers announce and on the certificates they send, which brings the way endorser blocks spread across the network closer to the design meant for production.

Two details in the report matter for anyone watching a test network. The prototype uses the Leios header format only from the Dijkstra era onward, Dijkstra being the name of the next Cardano era and hard fork. And the Leios prototype testnet has to be restarted from its genesis block, with node operators updating their software. Mainnet is not affected by that restart, the report says. The same week the team shipped ouroboros-consensus version 5.1.0.1.

Endorser-block selection is still missing

The report is just as clear about what is not done. Its next item is selecting and fetching endorser blocks. That is the heart of the scheme: in Leios, the ordinary blocks that stake pools produce about every 20 seconds can point to endorser blocks, which list extra transactions (by reference, not the transactions themselves, IOG explains), once pools holding at least 75% of active stake have voted to certify them. A node that cannot yet choose which endorser block to fetch, or retrieve it, is not running Leios end to end.

By the end of October, then, the protocol shell is merged, the messaging parts are in, and the formal specification has caught up with the Linear Leios design. The logic that decides which extra blocks a node actually takes is the next job, and the weekly report gives no target date for it.

Where the 15x figure comes from

The numbers come from IOG itself. In a post published on 20 October, “Leios is coming: How it reaches Cardano mainnet”, Ivan Irakoze, Ecosystem Communications Lead at Input Output, puts the current Praos ceiling at about 4.5 kilobytes of transactions per second (TxkB/s). The initial aim for Leios is three to five times that. After at least two later parameter updates, the post says, throughput could reach roughly 10 to 15 times.

IOG’s own warning is blunt: “The hard fork is the beginning of the rollout, not the point when Cardano suddenly runs at full Leios capacity.”

The same post cites an estimate from CIP-164, the Cardano Improvement Proposal that defines Linear Leios: fee income would need sustained traffic of about 50 to 70 TxkB/s to break even, which IOG describes as more than ten times what Praos can carry today. CIP-164 also states, in transactions per second rather than kilobytes, that stake smart contract holding reserves of two or more tokens, supplied by users, that a decentralized exchange uses to execute trades automatically at a price set by a formula.">pool profitability improves significantly once sustained throughput exceeds 30 to 50 TPS. The two units measure different things and cannot be converted without knowing the average transaction size, so we keep them apart.

The table below turns IOG’s multiples into kilobytes per second. The arithmetic is ours (4.5 multiplied by each multiple, and 50 to 70 divided by 4.5 for the last row); the multiples, the 4.5 TxkB/s baseline and the 50 to 70 TxkB/s band are IOG’s.

Read that way, the 10x step (45 TxkB/s) stays below the break-even band, and only the top of IOG’s range, 67.5 TxkB/s, falls inside it. And the range describes capacity, not demand: a chain that can carry 15 times more data still has to be used that much for the fee argument to work.

An older IOG article, from 17 April 2025, used a slightly different baseline of about 4.4 kB/s and said simulations suggested Leios could go past 1,000 TPS through parallel processing in the node. That was a simulation result, not a mainnet measurement, and it is not the figure IOG now puts forward for the rollout.

What is Cardano Leios, in short

Leios changes how much Cardano’s base chain can carry, not where transactions go. It is a change to the layer 1 itself, unlike rollups and payment channels that move activity off the main chain (our guide on the difference between layer 1 and layer 2 explains both approaches). Cardano keeps its proof-of-stake block producers and their roughly 20-second rhythm for ranking blocks; Leios adds endorser blocks that reference the extra transactions, plus votes by stake pools that certify them.

CIP-164, “Ouroboros Linear Leios – Greater transaction throughput”, was created on 7 March 2025 and still has the status Proposed in the CIP repository. Its authors include William Wolff, Brian W Bush, Sebastian Nagel and Nicolas Frisby, and it names Input Output Engineering as implementor. IOG’s 2025 explainer said the design keeps resistance to up to 50% adversarial stake.

Six steps between the merged code and mainnet

IOG’s 20 October post and Intersect’s weekly update 134 of 24 October together describe the road. None of the steps below has a calendar date.

1. A Dijkstra testnet that does not yet run Dijkstra rules

Intersect respun the public Dijkstra testnet on 21 October. It still runs protocol version 11, the current Conway era, on node 11.1.3. A hard fork initiation action to move it to protocol version 12, Dijkstra, will be submitted once a node 11.2 pre-release is available. The testnet uses one-day epochs, has a Kleioscan explorer and is open to stake pool operators now. Dijkstra rules themselves are not being tested yet.

2. The rest of the Leios code

Endorser-block selection and retrieval, as described above, are unfinished according to the 23 October report.

3. CAP-12, the constitutional amendment, in its fourth draft

CAP-12 would write into Cardano’s constitution the allowed ranges for Dijkstra-era parameters. On 19 October the chair of the Technical Steering Committee posted seven comments on behalf of the Parameter Committee. According to Intersect, he backed describing the Leios off switch as an orderly deactivation rather than an emergency one, asked for a written fast path for genuine Severity 1 or 2 incidents, passed on the advice to move Peras parameters into their own amendment, and swapped his earlier cap on how far pledge leverage can move in one step for a limit on how often it can change.

The author posted a fourth draft two days later. Peras is out of it, and it now lists the five constitutional provisions it changes. Still open, Intersect says: the pledge-leverage range, placeholder numbers, a short description of the off switch and whether CIP-23 ships in the first or second Dijkstra phase. The editors’ review also found a copy error, a new fast-path guardrail attached to the wrong parameter. Changing the constitution takes 75% of DRep stake and two-thirds of the Constitutional Committee; stake pools do not vote on it, according to IOG’s table.

4. The hard fork vote

The Dijkstra hard fork itself needs 60% of DRep stake, 51% of stake pool stake and two-thirds of the committee, IOG says. Leios would not travel alone. IOG lists nested transactions, account enhancements, guard scripts, Plutus V4, freely withdrawable staking rewards and a pledge-leverage parameter in the same bundle. Neither IOG nor Intersect gives a date for the fork.

5. BLS keys from 90 to 95% of stake

After the fork, each stake pool operator must generate a BLS key and submit an updated pool registration certificate that carries it; the key becomes active at the next epoch boundary. No vote is involved, and IOG says pools will rotate these keys roughly every quarter, on the same rhythm as their KES keys. In IOG’s words: “Leios needs roughly 90 to 95% of the stake registered before it switches on.” Intersect puts that at roughly the top 500 to 900 pools out of about 2,700. A Leios certificate then needs the votes of 75% of active stake. Pools without a key keep producing ordinary Praos blocks but cannot vote on endorser blocks.

6. Load tests, then two parameter votes

IOG plans mainnet load exercises before raising throughput: scheduled bursts of traffic large enough to overflow ordinary blocks, so that endorser blocks, votes and certificates are produced and timed on the real network. These need no vote. Then come at least two parameter updates, each needing 67% of DRep stake, 51% of pool stake and two-thirds of the committee. As long as the new values stay inside the limits the constitution sets, no further hard fork is needed. This is the stage at which the 10 to 15 times range could be reached.

Source: IOG, 20 October 2026. BLS registration is not a vote but has its own threshold: about 90 to 95% of delegated stake.

Is there a Cardano Leios release date?

No. IOG’s post, Intersect’s update and the consensus report give none for the Dijkstra hard fork, for BLS key registration or for the load tests. The only conditional timing in the sources is Intersect’s: the testnet hard fork action waits for the node 11.2 pre-release, which itself has no published date. Claims circulating in Cardano news roundups about which node version will carry Leios have not been confirmed by a primary source, and we do not repeat them here.

The order of events, though, is fixed by the rules. Code and testnet first, then the constitution amendment, then the fork vote, then key registration by pools, then load tests and parameter votes. The DReps and the committee who will vote on these steps are the same bodies that already approve Cardano’s treasury withdrawals, covered in our report on the Cardano treasury’s remaining 42.6M ADA under its spending cap.

What it changes for ADA holders and pool operators

For stake pool operators, the practical message is to join the Dijkstra testnet, which Intersect says is open to them now, and to expect a BLS key and certificate update after the fork. A pool that skips it still makes blocks but has no say in certifying endorser blocks, and Leios cannot switch on until 90 to 95% of stake has registered.

For DReps and the Constitutional Committee, the work is CAP-12, then the fork vote, then the throughput votes. Our reading of IOG’s post and Intersect’s update: none of the steps they list gives a task to people who simply hold ADA. Developers, pool operators, DReps and the committee do the work, and holders take part only through the pool and the DRep they delegate to.

ADA traded at about $0.2047 at 11:46 UTC on 26 October, up 0.63% over 24 hours and down 7.19% over seven days. Nothing in that data links the move to the Leios news.

Next dates to watch

• 27 October, 16:00 UTC: CIP Editors Meeting (Intersect).
• 29 October, 15:00 UTC: CAP review of the fourth draft of CAP-12 (Intersect).
• No date: the node 11.2 pre-release, which triggers the hard fork initiation action on the Dijkstra testnet.
• No date: endorser-block selection and fetching in the consensus code, named as the next task in the 23 October report.