Aptos has unveiled MonoMove, a redesigned execution engine for the Move virtual machine that the foundation says runs smart contracts up to 55 times faster for onchain markets. The figures come from internal benchmarks that replayed real mainnet transactions from DecibelTrade, and the engine is not scheduled to reach mainnet before 2027 — and only then if governance approves.

The claim targets a bottleneck that transactions-per-second marketing rarely addresses: how fast the virtual machine processes the code inside each transaction. MonoMove is a complete rebuild of the Move VM and its surrounding execution stack, not a tuning pass on the existing design. Aptos has already integrated the engine into its block executor, the component that processes transactions in each block, and development is proceeding through GitHub pull requests throughout 2026.

How does MonoMove reach those speeds?

The redesign pursues two goals: better single-thread performance on one processor core, and parallelism that lets jobs run side by side without conflicts. Two technical choices drive both.

Monomorphized micro-ops generate specialized versions of code ahead of time, so the engine avoids resolving data types on the fly while a transaction executes. A register-based interpreter holds values in named slots instead of constantly pushing and popping them from a stack. Together, the approaches strip wasted work from every instruction.

What do the benchmarks show?

Aptos replayed DecibelTrade transactions against the previous VM and reported results by transaction type:

• Collateral withdrawals: up to 55x faster
• Vault requests: up to 40x faster
• Perpetual market requests: up to 38x faster
• Order placements: up to 22x faster

End-to-end throughput improved 3-8x across workloads. Order-book matching gained roughly 7x, liquidity-pool swaps 5-7x, and lending markets were also part of the tested set.

The gap between the 55x headline and the 3-8x end-to-end figure reflects a structural reality: execution is only one stage of processing a transaction. Even so, the throughput improvement compares favorably with the roughly 2x gains Aptos achieved in 2025 with Loader V2, its previous performance upgrade.

What still has to be built?

MonoMove is not live for users, and the roadmap runs in stages — sequential execution first, then advanced parallel support. Several features remain unfinished:

• Native function ports
• Reentrancy checks, which guard against the classic exploit where a contract is called back into before its first job finishes
• Gas instrumentation, the metering that measures computation so users pay for what they consume

Aptos targets completion of the full feature set by the end of 2026. The company plans differential testing — feeding identical inputs to both engines and flagging mismatches — plus formal verification against the legacy AptosVM, using mathematical proofs to confirm the new code behaves identically. The redesign is also meant to stay within Move's resource-oriented safety model, which treats digital assets as resources that cannot be copied or accidentally destroyed, while adding runtime checks.

Why the benchmarks matter for onchain trading

The benchmark selection signals where Aptos is aiming. Collateral withdrawals, vault requests, perpetual market requests and order placements are the core plumbing of onchain derivatives and trading venues — venues where execution latency directly determines whether a market can compete with centralized exchanges.

The numbers deserve scrutiny. The 55x figure is an upper bound from internal tests on one application's replayed transactions, not an independent audit, and production performance could look different once gas metering and reentrancy protection add their own overhead. Feature completeness by end-2026 is a target, not a guarantee.

Observers have concrete milestones to track: the pace of MonoMove pull requests through 2026, whether parallel execution support lands on schedule, the results of differential testing against legacy AptosVM, and ultimately the governance proposal that would bring MonoMove to mainnet in 2027. That vote is the final checkpoint between the benchmarks and a production network.