Cardano Leios Moves From Prototype to High Confidence Testing

Cardano’s Ouroboros Leios project is moving into a more demanding phase of development, where the central question is no longer only whether the network can increase throughput, but whether that increase can be proven safe under real network conditions. The May 2026 Leios Monthly Review focused on prototype progress, governance support, red team testing, transaction submission stress tests and the engineering work needed before Leios can move closer to mainnet.

Leios is designed as a high throughput protocol for Cardano that increases transaction capacity while preserving security and ecosystem compatibility. The official Cardano update published in May stated that IOG submitted a ₳27.7 million treasury proposal to mature Ouroboros Leios from a testnet prototype into a mainnet ready release candidate by late 2026, with a phased throughput goal of 10x to 65x.

The May review adds a more technical layer to that roadmap. The team said the current proposal to continue shipping Leios toward mainnet had been voted in, ratified and was soon to be enacted, with support described as around 88%.

That puts Leios directly inside Cardano’s governance process, where protocol development, DRep voting and SPO confidence now have to move together.

Cardano governance gives Leios a path toward mainnet

The review showed continued progress on the Leios prototype, including chain visualization, voting logic and the certification flow around endorser blocks. In the demo, nodes produced blocks, announced endorser blocks, collected votes and reached the required threshold for certification before transactions could be reflected in the ledger state.

In practical terms, Leios adds a new layer of work around block production. Instead of relying only on standard blocks to carry transaction load, endorser blocks allow larger sets of transactions to be proposed, voted on and certified. Once a certificate is included, those transactions can be enacted in the ledger state.

The team also connected the prototype work to the Dijkstra era, BLS cryptography, key registration and future integration points for node implementations. That distinction matters. The prototype is being used as a staging and experimentation area, not as code that can simply be pushed into production without further engineering. The review made clear that additional production grade implementation work is still required before Leios can be treated as a mainnet upgrade candidate.

For Cardano, this is the important governance angle. Leios is not just a throughput upgrade being developed in isolation. It is a protocol change that will require technical evidence, implementation readiness and enough confidence from the operators who will eventually install and run the software.

Leios high confidence work introduces red team testing

The strongest part of the May review was the introduction of the high confidence workstream. The team framed this as a gating factor for Leios, because DReps first need to vote for a hard fork and SPOs then need enough confidence to install it. Mathematical proofs and development progress are not enough on their own. The network also needs evidence that the design behaves safely when conditions are adversarial or imperfect.

That is why a red team has started working with the Rust simulator, creating tools for deliberate attack scenarios. The goal is to validate the existing threat model by testing how Leios behaves when nodes fail, refuse to cooperate or try to disrupt expected behavior. These scenarios are expected to be tested in simulation and later on testnets.

A key tool in that process is Piranha, a light node without a ledger component but with real networking and consensus behavior. It can speak through real mini protocols and can be instructed to behave badly, allowing the team to coordinate failure or attack patterns across multiple nodes. That makes it possible to test how the rest of the network responds when some participants defect from expected behavior.

One early example was the lazy voter scenario, where a committee member that should vote refuses to do so. The team also described work on scenarios involving endorser block distribution failures and equivocation attacks, where a node may send different endorser block information in different directions.

This is where the Leios story becomes more than a scaling headline. The project is now testing whether the protocol can preserve safety and recover predictable behavior when real world failure modes are introduced. For a consensus upgrade that could reshape Cardano’s throughput profile, that work is as important as the throughput target itself.

Mempool stress tests show where Cardano scaling gets difficult

The review also showed that some of the hardest Leios work now sits in the mempool, transaction diffusion and network modeling layers. Stress tests used a less connected topology on purpose, with nodes spread across Europe, the United States and Asia Pacific. The test used a 25 megabyte mempool and was designed to put pressure on transaction submission and mempool behavior.

The results showed clear regional differences. European nodes, closer to the relay setup, filled their mempools faster. United States nodes took longer but eventually reached the target. Asia Pacific nodes, with higher latency, were not able to fully fill their mempools in the same way.

That finding does not weaken the Leios case. It identifies the layer that must be improved. The team said the current mempool is not optimized for this kind of high throughput scenario, while noting that CPU capacity was not the main bottleneck in comparable benchmark conditions. The challenge is the way transactions are diffused, validated, revalidated and connected to different ledger states.

Leios makes that problem more complex because the system has to reason about more than one ledger state. A transaction that is valid against the basic ledger state may not be valid against a state extended by endorser block inclusion, and the reverse can also be true. That creates open design questions around whether Leios needs different mempool views, how those views should be synchronized and how block forging can remain fast enough.

The team also discussed the late joiner problem. In a demo, all nodes can start together. On a real testnet or mainnet, nodes may join late, go offline and return after missing key messages. The proposed fix allows a node that missed a Leios block offer to request the missing endorser block data and catch up with the rest of the chain.

Network modeling was another major part of the review. The discussion covered peer selection, endorser block fetching, high latency links, slow response behavior, Delta Q modeling and TCP congestion algorithms such as Cubic, Reno and BBR. The detail is technical, but the message is clear: Leios is being tested against the real constraints of a global decentralized network, not only against clean laboratory assumptions.

That is why the May review is an important marker for Cardano. Leios is now being pushed through the part of protocol development that matters most before a hard fork, where throughput goals have to survive latency, mempool contention, node recovery, adversarial behavior and governance scrutiny. If the project reaches a mainnet ready candidate by the end of 2026, the value will not only be in higher capacity. It will be in the evidence that Cardano can scale its consensus layer through a process that operators, developers and governance participants can independently trust.