Smart Wallets Speed Comparison: Real Latency Numbers
If you have used a smart wallet for anything time-sensitive — a presale claim, an airdrop window, a liquidation defense — you have probably noticed it does not feel as snappy as a plain MetaMask transaction. This smart wallets speed comparison goes through what actually slows things down, with measured numbers from public bundler benchmarks, and where the speed difference matters versus where it does not.
We are not here to sell you on account abstraction. We are here to tell you what the extra plumbing costs you in seconds.
What “speed” actually means for a smart wallet
A normal externally owned account (EOA) signs a transaction and broadcasts it to a node. One hop. A smart wallet under ERC-4337 does something different: it builds a UserOperation, sends it to a bundler, the bundler simulates it, optionally a paymaster sponsors gas, and then the bundler submits a real transaction that calls the EntryPoint contract, which calls your wallet contract, which calls the target contract. Multiple hops. Multiple round trips. (Ethereum.org, accessed May 2026)
So when we say “speed,” we mean three different things:
- Sign-to-broadcast latency — how long from clicking “confirm” to the user op being publicly visible in the mempool or bundler queue.
- Broadcast-to-inclusion latency — how long until it is in a block.
- Perceived UX latency — including biometric prompts, passkey roundtrips, and any pre-flight simulation.
The third one is what users actually complain about, but it is the least standardized.
The numbers we have
Pimlico published bundler latency data in late 2024 showing median bundler-to-inclusion times of roughly 2-4 seconds on Ethereum mainnet, with Base and Arbitrum coming in under 1 second for the same flow. (Pimlico, 2024) Stackup’s public benchmarks showed similar ranges, with p99 latency on mainnet sometimes spiking past 12 seconds during congested periods. (Stackup docs, 2025)
Compare that to a vanilla EOA: if you set a competitive gas price, your transaction is in the mempool the instant your RPC accepts it, and inclusion is governed only by block time and your tip. There is no bundler in between.
So as a rough order of magnitude, on Ethereum mainnet, a smart wallet user op costs you roughly 1-4 extra seconds over an EOA in the median case, and significantly more during congestion.
How the major smart wallets compare
We will not invent numbers we did not measure. What we can say from public documentation and our own informal testing on Base mainnet:
Coinbase Smart Wallet. Uses passkeys (WebAuthn) for signing, runs through Coinbase’s own bundler infrastructure on Base. The passkey prompt itself is fast — sub-second on most devices — and the bundler path on Base is short. End-to-end from click to inclusion was consistently the fastest in our testing. (smartwallet.dev)
Safe (formerly Gnosis Safe). Safe is a multisig first, smart wallet second. If you have a 2-of-3 setup, your transaction is not going anywhere until two humans sign it. That is by design and it is the right design for treasury custody, but it is structurally slow for anything time-sensitive. Single-signer Safes with 4337 modules are competitive with other 4337 wallets but you lose the multisig benefit. (Safe docs)
Argent. Pioneered smart wallet UX. On StarkNet (Argent X) it is fast because the chain is fast. On Ethereum L1, it has the same bundler tax everyone else does. Their guardian-based recovery is excellent for security but does not affect transaction speed.
Biconomy / ZeroDev / Pimlico SDKs. These are infrastructure rather than user-facing wallets, but most consumer apps using account abstraction are routed through one of them. Speed depends entirely on which bundler the dApp picked and which chain you are on.
Where speed actually matters
Here is the retail-skeptical part. For most things you do, the bundler latency is irrelevant. Sending USDC to a friend? Two extra seconds is invisible. Voting in a DAO? Nobody cares.
Where it does matter:
- Sniping a public token launch. If you are competing with bots, you are already losing. A smart wallet just makes it slightly worse. Use an EOA or do not play.
- Presale claim windows. Some presale claim contracts have first-come-first-served allocations. We have a separate writeup on this in our guide to picking a wallet for presale claims.
- MEV-sensitive trades. Same issue. Smart wallets are not designed for being faster than searchers.
- Liquidation defense. If your CDP is about to be liquidated, every second matters. Most lending protocols allow EOA top-ups, and you should use one.
For everything else — ongoing portfolio management, multisig treasury, recovery-friendly storage — the speed difference is a non-issue and the security upgrades from account abstraction are worth it. We covered the security side in our custody options breakdown and discussed seed phrase risks here.
What we could not verify
We did not run controlled benchmarks across all wallets on the same hardware on the same day. Public bundler stats vary by who is publishing them. The Pimlico and Stackup numbers cited above are the operators’ own measurements, and you should treat them as motivated. If anyone runs an independent ERC-4337 latency benchmark with reproducible code, we will link it here.
We also do not have hard numbers for how often paymaster failures cause a user op to be silently dropped versus retried. Anecdotally, this happens enough that “the wallet just sat there” is a real complaint, not a misperception.
The actually-fast option for time-critical stuff
If you genuinely need the fastest possible inclusion on Ethereum L1 for a specific transaction — a presale buy, a claim, an arbitrage — use a fresh EOA, fund it with the exact amount needed plus gas, and broadcast through a private RPC like a Flashbots endpoint to avoid frontrunning. We discussed this approach in our presale safety checklist. It is uglier than a smart wallet, but it is honest about what it is doing.
Honest summary
Smart wallets are not slow in any meaningful way for normal use, and the security and recovery wins are real. They are measurably slower than EOAs for time-critical transactions on congested L1, by roughly 1-4 seconds in the median case and more at the tail. If you are doing presale sniping or competing with bots, account abstraction is not your friend; use an EOA. For everything else, pick a smart wallet on the basis of recovery model and bundler reliability, not raw speed, because the speed difference will not change your life.