Input Output has outlined a staged rollout for Cardano’s Leios upgrade, with an initial target of three to five times the network’s current transaction capacity before later governance votes could increase throughput to 10–15 times.
Leios will become available with the Dijkstra hard fork, but activation will require stake pool operators to register new BLS voting keys. Input Output estimates that voting keys representing about 90–95% of delegated stake must be registered before the system can operate.
“The hard fork makes the protocol available,” Input Output said in its 20 October update, which did not provide a fixed activation date for Dijkstra.
Stake pool operators will generate BLS keys after the hard fork and add them through updated pool certificates. Registrations become active at the next epoch boundary. BLS signatures allow hundreds of pool votes to be combined into a single compact certificate.
Leios requires votes representing at least 75% of active stake to certify an endorser block, which contains references to transactions waiting to be processed. Input Output says registration must be above that threshold to account for pools that are offline or slow to respond. It estimates participation from about 500–900 of the largest pools, although the final requirement depends on parameters still being set.
Pools without a BLS key can continue producing Praos blocks, but cannot vote on Leios endorser blocks. Registration does not require a separate vote from delegated representatives or the Constitutional Committee.
The rollout follows the 18 July van Rossem hard fork, which moved Cardano’s mainnet to Protocol Version 11. Intersect confirmed that upgrade after testing and approval by the network’s governance bodies.
Capacity increases will require votes
Input Output plans a constitutional update before Dijkstra to define the new protocol parameters and the limits within which governance can change them. The Intersect Parameters Committee is responsible for preparing the guardrails, while the Plutus team will update the guardrails script.
The constitutional amendment requires support from delegated representatives controlling 75% of stake and two-thirds of the Constitutional Committee. The hard fork requires 60% of DRep stake, 51% of stake pool operator stake and two-thirds of the committee. Later throughput increases require 67% of DRep stake, 51% of SPO stake and two-thirds of committee members.
After Leios goes live, controlled bursts of mainnet traffic will test block delivery, voting speed, certificate reliability and spare processing capacity on nodes. The first target is based on the current Praos ceiling of about 4.5 kilobytes of transactions per second, equivalent to approximately 13.5–22.5 TxkB/s. The later target would be 45–67.5 TxkB/s.
Those figures measure transaction data rather than a fixed number of transactions per second, since the number processed depends on transaction size. Input Output says limits can be changed without another hard fork if proposals remain within the approved constitutional guardrails.
Testing on MusashiNet reached 26.8 TxkB/s against a Praos ceiling of 4.51 TxkB/s, or about 5.94 times the capacity. The 41-day Earth phase processed more than 127,000 blocks with 63 registered pools, using synthetic traffic and participating operators. Problems including memory leaks, chain forks, stalled synchronisation and a network halt caused by an invalid certificate being treated as fatal were subsequently corrected.
Input Output plans conservative mainnet settings because Cardano has about 2,700 registered pools with varied hardware, connections and configurations. Leios retains Praos ranking blocks and adds endorser blocks when pending traffic exceeds the capacity of a ranking block. Transactions handled through endorser blocks are expected to reach the ledger within roughly 45–60 seconds.
A separate Cardano update said 21 of 103 listed apps had enabled transaction tracking, representing about 20.4% of projects. Input Output also said more than 85% of epoch rewards come from Cardano’s Reserve, while a CIP-164 estimate puts break-even fee income at sustained traffic of roughly 50–70 TxkB/s.
“Leios creates a route to that level of activity,” Input Output wrote. “It does not create demand by itself.”
Dijkstra will also include nested transactions, account enhancements, guard scripts, Plutus V4, freely withdrawable staking rewards and a pledge-leverage parameter. Linear Leios is planned for the first phase of Intersect’s roadmap, while Peras is targeted for the second quarter of 2027.
