Mysticeti v2 reduces latency for Sui’s Move-based object model
Mysticeti v2 reduces consensus latency to 390 milliseconds with deterministic finality for Sui. The update utilizes a new Transaction Driver and signature batching to improve performance while leveraging the Move language to prevent asset-level bugs and reentrancy attacks.
Sui settles in 390 milliseconds with deterministic finality. Mysticeti v2 reduces consensus latency through an uncertified DAG and multi-leader proposals. For Asia-based nodes, latency fell from 1.00 second to 0.65 seconds, while for Europe-based nodes, latency fell from 0.55 seconds to 0.40 seconds. Mysticeti v2 uses a new Transaction Driver where a client submits to a single chosen validator and retrieves effects with quorum validation, replacing the old Ququorum Driver to reduce redundant consensus work. Signature batching during block signing reduces CPU load. The protocol uses a universal commit rule to commit blocks as soon as conditions are met. Pipelining processes multiple waves of transactions simultaneously to keep commitments flowing through the network. Mysticeti introduces native randomness using threshold cryptography and Distributed Key Generation. Owned objects follow a fast path that settles without waiting for consensus. Shared objects, such as a DeFi pool, follow the full consensus path. Mysticeti v2 and the Transaction Driver became the default from Sui node v1.60. This update narrowed the gap between the consensus path and the fast path for owned objects.
Move protects digital assets
Move prevents asset-level bugs by treating assets as resources. Sam Blackshear designed Move to ensure assets cannot be copied or accidentally destroyed. This design prevents reentrancy attacks because ownership transfers remain explicit and atomic. Bytecode verification checks for type safety and memory safety before deployment. If you know how to code, you see how this replaces manual checks. Five of the OWASP Top 10 vulnerabilities are impossible in Sui Move. Programmable transaction blocks allow developers to batch up to 1,024 operations. These blocks execute mass NFT minting or complex DeFi workflows atomically. Objects include address ownership, object ownership, shared ownership, and immutable objects. Address ownership includes NFTs or fungible tokens. Object ownership allows complex hierarchies where one object owns another. Shared ownership allows multiple entities to access an object. Immutable objects are read-only and cannot be modified once created. Linear logic ensures that resources cannot be duplicated or modified by unauthorized code. Sui Move functions as a derivative of the Move language originally developed at Mysten Labs. Developers use the Move Prover for formal verification. The move-analyzer plugin for Visual Studio Code provides real-time feedback.
Throughput and the ecosystem gap
The object model enables high per-account throughput because transactions only contend when they touch the same object. This architecture differentiates Sui from Solana’s Sealevel, which relies on an account model. Solana’s Alpenglow upgrade targets 100 to 150 millisecond finality, but the upgrade lowers the adversarial stake threshold to 20 percent. Sui holds $473 million in DeFi TVL. This amount is significantly lower than Solana’s $5.9 billion.
| Metric | Sui | Solana (Alpenglow) |
|---|---|---|
| Deterministic Finality | 390 ms | 100 – 150 ms |
| DeFi TVL (Sept 2026) | $473 million | $5.9 billion |
| Adversarial Stake Limit | 33% (traditional) | 20% |
Sui provides an integrated stack including Walrus for storage and Seal for data encryption. It includes DeepBook for liquidity and Nautilus for off-chain computation. Users use existing Google or Apple accounts to sign in via zkLogin. Solana’s account model requires transactions to declare the accounts they touch, which leads to bottlenecks when many transactions modify the same account. Solana treats NFTs as metadata inside accounts. Sui treats them as distinct objects with intrinsic properties and permissions. Validators process transactions individually to reduce network latency. Users pay gas fees consisting of computation and storage components. Users can delete stored data to reclaim fees from the storage fund. Sui’s scalability is horizontal and theoretically uncapped because validators increase capacity by adding worker machines to their infrastructure. Will the object model eventually close the massive TVL gap between Sui and Solana?
Join the discussion