An externally-owned account (EOA) is a keypair. A smart wallet is a contract account. That single distinction at the protocol level drives every other difference: signing model, recovery story, gas mechanics, programmability, and audit risk. ERC-7702, which went live in the Pectra upgrade on May 7, 2025, partially collapses the gap by letting an EOA temporarily execute contract code, but the underlying split between keypair accounts and contract accounts remains the right mental model.
This guide compares the two account types across the dimensions that matter when picking one. It also walks through the four most common user profiles, the real gas overhead of a smart-wallet transaction, the myths worth ignoring, and the migration paths from a plain EOA to Safe or to an ERC-7702 delegation.
The short version: keypair account vs contract account
An EOA is controlled by a single private key. The Ethereum protocol checks an ECDSA signature against the address, and if the math works, the transaction executes. There is no code at the address, no logic to evaluate, and no way to add features. Recovery means whoever holds the seed phrase controls the funds. The model has been unchanged since 2015.
A smart wallet is a smart contract deployed at an address. Authorization rules live in the contract code. The wallet can require one signature, multiple signatures, a passkey, a session key with limits, a recovery threshold, or any combination. Anything the EVM can express, the wallet can enforce. The cost is a deploy transaction the first time, plus a small per-transaction overhead because every call routes through the contract.
Smart wallet vs EOA at a glance
The table below summarizes the nine dimensions that matter when choosing between an EOA and a smart wallet. Every row reflects how the EVM treats each account type, not marketing positioning.
Dimension | EOA | Smart wallet |
Signing model | Single ECDSA key (secp256k1) | Arbitrary: ECDSA, passkey (WebAuthn / secp256r1), multisig, session keys |
Gas payment | Native token only, paid by sender | Native token, ERC-20 tokens, or sponsored via paymaster |
Recovery | Seed phrase only. Lose it, lose the funds. | Social recovery, multisig threshold, passkey backup, time-locked guardian rotation |
Programmability | None. The account is the key. | Full: batched calls, conditional logic, hooks, plugins (ERC-6900, ERC-7579) |
Per-tx fee | Base 21,000 gas plus call data | 21,000 base plus ~30k to ~100k contract overhead per UserOp |
Audit risk | Zero contract surface. Risk is key custody only. | Contract risk. Mitigated by audited implementations (Safe, Coinbase, Argent) |
Deploy cost | None. Address is derived from the key. | One-time deploy (~100k to 400k gas) or counterfactual deploy on first use |
Multi-sig | Not possible at protocol level | Native: 2-of-3, 3-of-5, any quorum the contract specifies |
Session keys | Not possible | Yes, via ERC-7715 or implementation-specific permission modules |
What does ERC-7702 change about this comparison?
ERC-7702 is the most important account-abstraction change since ERC-4337. It lets an EOA temporarily point at a smart-contract implementation for the duration of a transaction. The address stays the same, the private key still works, but during the call the EOA executes whatever bytecode it delegated to. After the transaction, it returns to behaving like a regular EOA unless the delegation is left in place.
The practical effect: an EOA holder can batch calls, sponsor gas through a paymaster, or expose a session-key permission system without giving up the EOA address or migrating funds. The split between "EOA" and "smart wallet" still exists at the protocol level, because ERC-7702 is a delegation, not a transformation. But for most user-facing features, the gap is closing. Wallet vendors are shipping ERC-7702-flavored upgrades to MetaMask, Trust Wallet, and Rainbow that give EOAs smart-wallet ergonomics without forcing a fresh deploy.
The EIP-7702 specification is the authoritative reference. It shipped as part of the Pectra hard fork on May 7, 2025.
Real gas overhead of a smart-wallet transaction
The most common objection to smart wallets is fees. The objection is fair but smaller than people assume. A plain ERC-20 transfer from an EOA costs roughly 21,000 gas for the base transaction plus 30,000 to 50,000 gas for the token transfer itself. The same transfer routed through a Safe or an ERC-4337 smart account adds the validation step inside the contract, plus the bundler's call overhead if using ERC-4337.
Concrete numbers from Safe and ERC-4337 reference implementations:
Safe 1-of-1 owner transaction: roughly 50,000 to 80,000 gas overhead above the equivalent EOA call. A typical USDC transfer that costs ~65,000 gas from an EOA costs ~115,000 to 145,000 from a Safe.
ERC-4337 UserOperation: 80,000 to 120,000 gas overhead for validation, paymaster check, and bundler economics. The full transaction usually settles in the 150k to 250k gas range for simple calls.
ERC-7702 delegated call: closer to the EOA baseline. The delegation itself is set once via SET_CODE; subsequent calls execute the implementation directly with minimal extra overhead beyond the implementation's own logic.
On L2s, where gas pricing is dominated by data rather than execution, the percentage overhead is smaller. A Safe transfer on Base or Arbitrum costs cents, not dollars. The overhead matters more on Ethereum mainnet, where 50k extra gas at 30 gwei is roughly $4 at $2,500 ETH.
When to pick an EOA
EOAs remain the right answer in several common cases. They are not legacy technology to be replaced.
DeFi power user holding the keys yourself. If you understand seed-phrase hygiene, want zero contract-risk exposure, and value gas efficiency on mainnet, an EOA with a hardware wallet is hard to beat. Every DeFi protocol speaks EOA natively.
Operational accounts that need maximum portability. An EOA address works identically on every EVM chain without any deployment. A smart wallet's contract code may not exist on a less common chain yet, or may exist at a different address.
Bots and automated trading. Programmatic systems that sign their own transactions usually do not need recovery, multisig, or paymasters. The simpler the account, the easier the operational stack.
One-off receive addresses. Cold storage, deposit addresses for centralized exchanges, and similar receive-only roles do not benefit from contract logic.
When to pick a smart wallet
Smart wallets are the right answer wherever the EOA's "single key = full control" model is operationally painful or actually dangerous.
App builders shipping to retail users. Asking a new user to write down twelve words is a known dropoff cliff. Coinbase Smart Wallet, Argent, and similar passkey-based smart wallets remove the seed phrase entirely. Onboarding completion rates improve materially.
Treasury holders and DAOs. Safe is the default. A 3-of-5 multisig with explicit signer rotation is structurally safer than a single key held by one team member. Tens of billions of dollars sit in Safe contracts today for this reason.
Power users who want batching and session keys. Approving and swapping in one transaction, or granting a game a session key that can only spend up to $50 of a specific token for the next hour, are smart-wallet-only features.
Anyone who wants to recover from a lost device. Social recovery, where a configurable set of guardians can rotate the signing key, only exists in a contract account. EOAs have no recovery story beyond the seed phrase.
Common myths worth ignoring
A few claims travel further than they should. The corrections matter when deciding between account types.
"Smart wallets are too expensive to use." True on Ethereum mainnet for sub-$100 transactions, false on every L2. Base, Arbitrum, Optimism, and Polygon all push smart-wallet transaction costs into the cents range. Where most retail activity already happens, the gas argument has evaporated.
"Smart wallets are not real wallets, they are just contracts." They are contracts, and that is the point. Safe holds more dollar value than most banks. Calling that "not a real wallet" is a category error.
"You cannot use smart wallets across chains." Smart wallets deploy per chain, but most major implementations (Safe, Coinbase Smart Wallet, Argent) use deterministic addresses via CREATE2 so the same address works on every supported chain. ERC-7702 sidesteps the deploy entirely by letting an existing EOA delegate.
"ERC-4337 replaced EOAs." No. ERC-4337 added a parallel transaction format (UserOperations) that smart wallets can use. EOAs still exist, still work, and still account for the majority of active addresses. ERC-7702 is the protocol-level change that actually narrows the gap.
Migration paths: EOA to Safe, EOA to ERC-7702 delegation
Two clean migration paths exist for an EOA holder who wants smart-wallet features.
EOA to Safe. Deploy a new Safe through app.safe.global, set the signers and threshold (you can include the original EOA as one of the signers), then transfer assets from the EOA to the Safe address. The original EOA still exists; it has just been demoted to a signer of a larger account. This is the standard treasury migration. Costs: a deploy transaction (one-time, ~300k gas) plus the transfer of each asset.
EOA to ERC-7702 delegation. Sign a 7702 authorization pointing the EOA at a smart-account implementation (the wallet vendor provides one). After Pectra, the EOA can execute that implementation's code while keeping its address and private key. Funds do not move. Cost: a single SET_CODE transaction, then standard execution costs. The delegation can be revoked at any time by signing a new authorization that points the EOA at the zero address.
The two paths solve different problems. Safe is the right answer when you want multisig, separation of duties, or organizational controls. ERC-7702 is the right answer when you want smart-wallet features (batching, session keys, paymasters) without changing the address or migrating assets.
What does a hybrid setup look like?
Most serious users end up holding both. A typical setup:
A hardware-secured EOA for cold storage. Mainly receive-only, signs occasional large transfers. Maximum portability, minimum contract surface.
A Safe multisig for treasury or business funds. 2-of-3 or 3-of-5 with explicit signers, used for the funds that would be catastrophic to lose to a single compromised key.
A passkey-based smart wallet for daily activity. Coinbase Smart Wallet, Argent, or a Trust Wallet smart account for swaps, mints, and apps. Sponsored gas via paymaster makes UX feel like Web2.
Optionally, an ERC-7702 delegation on the EOA used for daily DeFi. Adds batching and paymaster support to the existing EOA without moving funds.
The point is not which account type wins. The point is matching the account to the threat model and the operational tempo of the funds it holds.
Methodology and sources
Gas figures reflect mainnet measurements from Safe 1.4.1 transactions and reference ERC-4337 implementations (Alchemy, Pimlico, ZeroDev bundler benchmarks) observed in 2025 and early 2026. EOA baselines are the EVM spec constants. ERC-7702 mechanics are pulled directly from the EIP-7702 specification, which was included in the Pectra hard fork that activated on Ethereum mainnet at slot 11649024 on May 7, 2025. Safe deployment and signer model reflect Safe's public documentation at docs.safe.global. Coinbase Smart Wallet behavior reflects the June 2024 public launch documentation.
Primary sources: