Cardano Leios Testnet Reaches Six Times Mainnet Throughput
The completed Earth phase produced Leios’ first public throughput measurement under distributed network conditions. Water now moves into parameter testing, a redesigned mempool and broader stake pool operator participation.
By SongMarketCap
Updated:
Cardano’s public Leios testnet reached 26.8 transaction kilobytes per second during its completed Earth phase, according to Leios product manager Carlos Lopez de Lara. The observed rate was almost six times the current Cardano mainnet limit of approximately 4.5 TkB/s.
The result emerged during simultaneous load from Input Output and community participants rather than a formal maximum-throughput test. MusashiNet has now entered the Water phase as the team prepares different protocol configurations and the governance path required for Leios to enter Cardano through the Dijkstra hard fork.
Leios Earth Phase Delivers Its First Public Throughput Result
Ouroboros Leios is Cardano’s proposed base-layer scaling protocol. It operates alongside Ouroboros Praos and introduces endorser blocks, allowing larger volumes of transaction data to be processed in parallel while Praos remains the network’s underlying security and settlement mechanism.
MusashiNet is the public environment used to move the Leios prototype from controlled development networks into infrastructure exposed to real latency, hardware and network conditions. The Earth phase focused on verifying whether the complete protocol design continued to function in that setting.
Input Output used a transaction generator called TX Centrifuge to create enough traffic to activate Leios. Community participants simultaneously used TX Firehose, a simpler load-generation tool made available for public testing. The reported throughput appeared during one period when both sources were generating transactions.
The measurement is not a TPS figure. TkB/s represents the volume of transaction data processed per second, while the number of transactions varies according to their size and complexity. Because the run was not designed as a formal stress test, the observed rate does not establish a published capacity limit for Leios.
The network included 62 block producers during the Earth phase, while the development team released ten versions of the prototype over six weeks. Lopez de Lara also reported five incidents, including forks and testnet halts.
According to his update, each incident was traced to an implementation bug that had not appeared in smaller local environments. None was identified by the team as a fundamental failure of the Leios protocol design. The issues were analyzed, corrected and followed by updated prototype releases.
A dedicated red team is also testing Leios against its threat model, attacking the system to identify weaknesses before the prototype is converted into production-grade code.
Water Phase Adds Parameter Testing and SPO Rewards
The Water phase shifts the focus from validating the basic protocol design to exploring different configurations. The team will adjust timing settings, block sizes and component implementations to determine which combinations provide stable and sustainable throughput.
Testing will include a redesigned mempool intended to address limitations that become more visible under heavier transaction loads. The new component will be evaluated alongside further increases in testnet activity.
Input Output has also opened the Musashi Testnet Rewards Program for stake pool operators. The program rewards operators for maintaining infrastructure on MusashiNet and sharing operational data from their nodes.
The requested telemetry is expected to include forging, voting and diffusion logs, node versions, clock offsets and host resources. Combining information from multiple operators allows the team to reconstruct how blocks, votes and certificates move through a geographically distributed network.
The program supports up to 100 pools. Operators must demonstrate control of their pool, register a BLS verification key and provide a mainnet ADA address for payments. The testnet faucet delegates one million test ADA to each registered pool so it can produce blocks and participate in Leios voting committees.
Applications are open, although the detailed monthly qualification criteria are still being finalized. The official documentation states that rewards are based on participation and data sharing rather than technical performance.
Dijkstra Rollout Creates a Governance Dependency
Lopez de Lara confirmed that Leios is planned for deployment through the Dijkstra era, Cardano’s next hard fork. Before activation, the team intends to propose a technical update to the Cardano Constitution covering the new protocol parameters and their guardrails.
The update is expected to be submitted between late September and early October. Lopez de Lara said it should remain limited to technical changes required by Dijkstra, without combining them with unrelated constitutional amendments.
The rollout plan also calls for conservative initial Leios settings, followed by gradual capacity increases as operational evidence becomes available. Some parameters can be raised later but could not be reduced without another hard fork, while higher infrastructure requirements would directly increase costs for stake pool operators.
That schedule intersects with Cardano’s Update Constitutional Committee action. Four of the committee’s seven seats expire on September 1. If the replacement action is not ratified, the committee would fall below the minimum of five members required for normal governance operations.
An undersized committee would prevent the planned technical constitutional update from advancing until the committee is restored. Lopez de Lara therefore called on DReps and SPOs to vote on the current action, directly connecting the committee decision with the proposed Leios timeline.
Earth closes with a functioning public prototype, ten releases and no fundamental protocol failure identified by the team. Water must now convert the initial throughput result into a stable mainnet configuration, while the committee vote and technical Constitution update determine whether those parameters can enter Dijkstra governance on the proposed schedule.