Arbitrum vs Optimism vs Base: Top Ethereum L2s Compared

Introduction

Base moves about 94 transactions a second while Arbitrum One handles 14, yet Arbitrum One is the only one of the three whose regular contract upgrades wait 17 days (L2BEAT, 2026). All three are optimistic rollups, layer 2 (L2) networks that post their data to Ethereum ↗, and all three sit at L2BEAT Stage 1. They still differ in who writes the software, how a withdrawal is proven, who can change the rules and whether a token exists. This comparison sets Arbitrum One, OP Mainnet and Base side by side on stack, proofs, sequencers, governance, fees and activity, using dated L2BEAT (renamed ETHBEAT on 4 November 2026), growthepie and CoinPaprika figures from November 2026, which change between visits. It ends with a table that matches each chain to a priority.

How Do Arbitrum, OP Mainnet and Base Differ in Their Technology Stacks?

Arbitrum One, OP Mainnet and Base are optimistic rollups, layer 2 (L2) networks that post their data to Ethereum, yet each runs different software under different owners (L2BEAT, 2026). Arbitrum uses its own Nitro stack, OP Mainnet uses the OP Stack, and Base has run on its own base/base stack since 28 May 2026 (L2BEAT, 2026; Base docs, 2026).

Arbitrum One: Nitro and BoLD

Arbitrum One runs Arbitrum Nitro, software from Offchain Labs that wraps a modified Geth Ethereum client and adds a WebAssembly virtual machine, so contracts compiled to WebAssembly run beside Solidity ones (L2BEAT, 2026). The Arbitrum decentralized autonomous organization (DAO), a body of ARB token holders, governs the chain. Nitro's dispute system, BoLD (bounded liquidity delay), went live on 12 February 2025. Anyone with enough funds can challenge a bad state claim, and a last-second dispute delays confirmation by 14 days at most (Arbitrum docs, 2026; Offchain Labs, 2025). A proposer stakes 3,600 ETH on each assertion, and challengers post bonds of 555 WETH or 79 WETH, depending on the dispute level (L2BEAT, 2026; Arbitrum docs, 2026).

OP Mainnet: the OP Stack and the Superchain

OP Mainnet is the flagship chain of the OP Stack, open-source software that also runs Unichain, Ink and Soneium. L2BEAT labels a chain part of the Superchain when it appears in the Superchain registry or uses the shared Superchain configuration (L2BEAT, 2026). Upgrade 20, executed on 24 September 2026, moved OP Mainnet, Ink, Soneium and Unichain to Super Root dispute games. That format is built for cross-chain messaging, which the upgrade itself does not switch on. It followed Upgrade 18, which added the Cannon and Kona game type, and Upgrade 16, which made the OptimismPortal contract interop-ready (Optimism docs, 2026).

Base: from the OP Stack to its own stack

Base announced on 18 February 2026 that it would leave the OP Stack for a Base-operated stack ↗, base/base, built on the Reth Ethereum client (Base, 2026). The Azul hard fork made the switch on 28 May 2026, after which only base-reth-node and base-consensus nodes can follow the chain. Azul also set the chain on a path to 1 gigagas a second (Base docs, 2026). Base keeps the OP Stack specification and stays an OP Enterprise client, Optimism's paid support service. L2BEAT still tags Base "Built on OP Stack" on 5 November 2026 because the lineage and the specification remain, although node operators now follow base/base releases instead of Optimism's.

Data current as of November 2026.

Stack ownership decides who ships upgrades; the proof system decides how Ethereum learns whether a state claim is true.

How Does Each Chain Prove Its State, and Which L2BEAT Stage Has It Reached?

All three chains sit at L2BEAT Stage 1, yet each proves its state differently and releases withdrawals on a different clock. The proof design sets the wait, and the stage shows how much power a Security Council keeps over user funds (L2BEAT, 2026).

How each chain proves its state

A proof system lets Ethereum judge a disputed claim ↗ about a chain's state. L2BEAT labels Arbitrum One and OP Mainnet "Fraud proofs (INT)", interactive proofs that resolve over several transactions, and labels Base "Fraud proofs (1R, ZK)", single-round proofs backed by zero-knowledge (ZK) proofs (L2BEAT, 2026). Base's Azul upgrade lets a trusted-execution-environment (TEE) proof and a ZK proof back the same proposal, and a ZK proof can challenge a TEE-backed claim (Base docs, 2026). About every hour a validator posts an assertion on Ethereum that attests to a batch of Arbitrum transactions (Arbitrum docs, 2026). OP Mainnet withdrawals now run through Super Root dispute games, but no ZK proofs are live: Optimism's notes place ZK dispute games in a later upgrade (L2BEAT, 2026; Optimism docs, 2026).

How long each withdrawal to Ethereum takes

Only withdrawals through the official bridge wait out a dispute window; ordinary transactions inside each chain do not. Arbitrum One releases funds after about 6.4 days, the length of its 45,818-block challenge period (Arbitrum docs, 2026). OP Mainnet's Standard Bridge waits seven days (Optimism docs, 2026). Base waits five days for a single-proof game since the Beryl upgrade on 25 June 2026, and one day when a TEE proof and a ZK proof back the same proposal (Base docs, 2026). A challenge backed by a ZK proof pushes Base's resolution out to seven days. Beryl's shorter window frees capital for fast-bridge liquidity providers, and third-party bridges that use their own liquidity release funds sooner.

What Stage 1 means and what Stage 2 would need

Stage 1 means only a Security Council with at least 75% agreement, or a bug, can block or forge messages from the L2 to Ethereum, and users can exit if the council disappears (L2BEAT, 2025). Arbitrum One, OP Mainnet and Base pass that walkaway test on 5 November 2026. Stage 2 needs permissionless proofs, an exit window of at least 30 days and a council limited to bugs provable onchain. L2BEAT lists the same two gaps for all three: some upgrades leave less than 30 days to exit, and council powers extend beyond provable bugs. L2BEAT adds that stages measure decentralization, not protection from bugs.

Stage 1 describes what happens if an operator misbehaves; it says less about the operator's daily powers over ordering and upgrades.

Who Orders Transactions on Each Chain, and Who Can Change the Rules?

Each chain runs one centralized sequencer, the operator that orders and executes transactions, and each lets users bypass it through Ethereum. The differences lie in how long that bypass takes and how long the contracts must wait before they can change (L2BEAT, 2026).

One sequencer each and three force-inclusion delays

L2BEAT lists a single centralized sequencer on Arbitrum One, OP Mainnet and Base, and its summaries for Arbitrum One and OP Mainnet name sequencer front-running as a maximal extractable value (MEV) risk (L2BEAT, 2026). If the sequencer fails, a user sends the transaction to a contract on Ethereum and the chain must include it. On Arbitrum One that contract is the SequencerInbox; on OP Mainnet and Base it is the OptimismPortal. The path takes up to one day on Arbitrum One and up to 12 hours on OP Mainnet and Base. All three also let anyone propose state roots, so a failed proposer cannot freeze withdrawals.

Who can upgrade the contracts and how fast

Arbitrum One alone has a delayed regular upgrade path: 17 days 8 hours, made of an eight-day timelock on the L2, a 6-day 8-hour outbox wait and a three-day Ethereum timelock, which leaves users a 10-day exit window (L2BEAT, 2026). Its Security Council of 12 members, nine of whom must agree, can upgrade at once on an emergency path, and used it on 21 April 2026 for the KelpDAO exploit freeze and on 24 May 2026 for a bridge patch. OP Mainnet upgrades need both its Security Council and the Optimism Foundation, with no delay. Base upgrades need Coinbase's three-of-six Base Coordinator Multisig and the Base Security Council, also with no delay. L2BEAT counts the council and Coinbase's multisig together as nine of 12 entities, or 75%.

Data current as of November 2026.

How Arbitrum now sells transaction priority

Arbitrum replaced its Timeboost express-lane auction with Priority Gas Auctions on Arbitrum One, which L2BEAT dates to 23 September 2026. Timeboost, live since April 2025, auctioned a 60-second express lane (Arbitrum docs, 2026). Under the new rule a sender attaches a priority fee to a transaction, and the sequencer ranks waiting transactions in 125-millisecond rounds, two per 250-millisecond block (Arbitrum Foundation, 2026). Proceeds split 97% to the DAO treasury and 3% to the Arbitrum Developer Guild. A paid Fast Feed streams ordering data but does not change a transaction's place. The sequencer stays a single trusted operator that applies the rule.

Control over ordering and upgrades is one half of a chain's governance; the other half is who holds a vote.

What Do ARB and OP Do, and Why Does Base Have No Token?

ARB and OP are governance tokens, not fee tokens, because users pay gas in ETH on all three chains. Both sit far below their peaks: ARB 93.2% and OP 97.7% under their all-time highs on 5 November 2026 (CoinPaprika API, 2026). Base has no token at all.

ARB: a governance token with a DAO treasury

ARB has a total supply of 10 billion, of which about 6.89 billion circulate, and its all-time high of $2.40 came on 12 January 2024 (CoinPaprika API, 2026). Holders vote in the Arbitrum DAO, and L2BEAT lists Arbitrum One as governed by token voting (L2BEAT, 2026). Sequencing revenue goes to the DAO treasury: Timeboost proceeds did, and priority-fee proceeds now follow the 97%-to-treasury split (Arbitrum forum, 2026). That stream is shrinking: Timeboost's quarterly revenue fell from $2.22M in Q4 2025 to $184K in Q3 2026 (DefiLlama, 2026). Offchain Labs raised the Timeboost reserve price from 0.001 to 0.0075 ETH in February 2026 in response to observed market behavior (Arbitrum forum, 2026). Token holders decide by vote how the treasury is spent.

OP: governance plus a Superchain buyback

OP has a total supply of 4,294,967,296, of which about 2.29 billion circulate (CoinPaprika API, 2026). Governance has two houses: the Token House, where OP holders vote with token-weighted ballots, and the Citizens' House, where each member has one vote (Optimism docs, 2026). An approved 12-month program that began in February 2026 sends 50% of incoming Superchain revenue to OP buybacks. Optimism collected 5,868 ETH of revenue in the year to January 2026, which would have meant about 2,700 ETH, roughly $8M, of buybacks (Optimism governance forum, 2026). The purchases run over the counter and are reported publicly. Base's old revenue share ended, and Base is not entitled to the unvested part of its 2023 OP grant.

Base: no token and ETH for gas

Base has no network token. Jesse Pollak said on 15 September 2025 that Base would explore one and called the plan early (CoinDesk, 2025). No token had launched by 5 November 2026, and L2BEAT lists ETH as Base's gas token (L2BEAT, 2026). Base's governance therefore rests on the two-party upgrade approval described above, and a future token would add a new step to it. Optimism's governance still applies a 9% voting cap to Base (Optimism governance forum, 2026). B20, the standard behind Base's tokenized stocks, governs assets issued on Base and is not a network token (Base docs, 2026). Neither Base nor Coinbase has published a token date.

The tokens' weak charts and Base's missing one have no effect on what a user pays; the fee tables show that directly.

Which Chain Is Cheapest for a Transfer or a Swap Right Now?

All three charge cents or less, so fee gaps rarely decide the choice. In growthepie's hourly bucket for 09:00 UTC on 5 November 2026 the median token swap cost $0.0105 on Arbitrum One, $0.0034 on Base and under $0.0001 on OP Mainnet, against $0.0610 on Ethereum (growthepie, 2026).

Transfer and swap costs measured on 5 November 2026

The same growthepie bucket priced a native ETH transfer at $0.0067 on Ethereum, $0.0014 on Arbitrum One, $0.0003 on Base and $0.0001 on OP Mainnet. The median transaction cost $0.0231 on Ethereum, $0.0043 on Arbitrum One, $0.0020 on Base and under $0.0001 on OP Mainnet. A swap on Base cost about 6% of the Ethereum price and one on Arbitrum One about 17%. A direct quote at 10:53 UTC for a 21,000-gas ETH transfer gave $0.0037 on Ethereum, $0.0010 on Arbitrum One, $0.00031 on Base and $0.000054 on OP Mainnet (public remote procedure call (RPC) endpoints, 2026). Readers who want the mechanics can follow how Ethereum gas fees work and when they spike ↗. These are the same hour and quotes used in the companion explainer on L2 fees ↗.

Why the three fee levels differ

An L2 fee is the chain's own gas plus the cost of posting data to Ethereum. In the 10:53 UTC quote the Ethereum data part was under 1% of Base's transfer fee and about 6% of OP Mainnet's (public RPC endpoints, 2026). Arbitrum One's quote came from an estimate that does not split the two parts. Each chain also sets its own gas price, so a lower data bill does not guarantee a lower total. Gaps of a fraction of a cent rarely decide a choice; exit time, upgrade control and liquidity do.

Data current as of November 2026.

Cheap fees mean little if no one is there to trade, so activity and value secured come next.

How Do Activity and Value Secured Compare Across the Three Chains?

On 5 November 2026 Base secured $14.8B, Arbitrum One $9.6B and OP Mainnet $1.6B, and over the seven days to 4 November Base processed about 94 transactions a second against 14 on Arbitrum One and 13 on OP Mainnet (L2BEAT, 2026). Base leads both measures, and neither measure rates a chain's safety.

Value secured and what it does not measure

L2BEAT's value secured adds canonically bridged tokens, externally bridged tokens and natively minted tokens. It is not total value locked (TVL), the DefiLlama measure of assets sitting inside protocols. On 5 November 2026 Base's $14.8B was 57% of the three chains' $26.0B, and Base and Arbitrum One together held about 63% of all L2 value secured ↗ (L2BEAT, 2026). On 5 November L2BEAT tagged 39.5% of Arbitrum One's value, 34.4% of Base's and 19.2% of OP Mainnet's as externally bridged, which relies on third-party bridges. L2BEAT also tags Arbitrum One as a host chain for layer 3 (L3) networks, which settle on it instead of on Ethereum. It counted about $1.14B of Arbitrum One's value as ARB itself and $0.26B of OP Mainnet's as OP (L2BEAT, 2026).

Activity over the week to 4 November 2026

Over the seven days to 4 November 2026 Base averaged about 94 transactions a second, Arbitrum One 14 and OP Mainnet 13, against about 21 on Ethereum itself (L2BEAT, 2026). Base therefore ran roughly four and a half times Ethereum's own rate and more than six times Arbitrum One's. A transaction count includes automated activity and says nothing about value moved, so Base's lead in count does not translate into a lead in capital per transaction. Arbitrum One holds 37% of the three chains' value but runs 12% of their transactions, and OP Mainnet holds 6% of the value with 11% of the transactions.

Data current as of November 2026.

Size and traffic describe the present; the sections that follow turn the measured differences into a choice.

Which Chain Fits Which User, Builder or Chain Launcher?

The measured gaps are small for fees, larger for exit time and upgrade control, and largest for liquidity. A user weighs cost and exit, a developer weighs stack and governance, and a chain launcher weighs who owns the software. No single chain wins every axis.

Users and traders

OP Mainnet had the lowest median fees in the 09:00 UTC bucket on 5 November 2026, and Base held the most value, $14.8B (growthepie, 2026; L2BEAT, 2026). Base also offers the shortest official withdrawal, one day on the dual-proof path (Base docs, 2026). Arbitrum One is the only chain of the three where a regular contract upgrade cannot land without a 10-day exit window, which suits a user who wants notice before the rules change (L2BEAT, 2026). Moving funds quickly between the three relies on third-party bridges, not the official ones, and each hop adds that bridge's own risk. OP Mainnet and Base can change contracts at once, so their users rely on the Security Council and the Foundation or Coinbase's multisig, and L2BEAT lists the exit window as none for both.

Developers choosing a stack

All three run the Ethereum Virtual Machine (EVM), so Solidity contracts deploy on each, and Arbitrum One adds a WebAssembly virtual machine (L2BEAT, 2026). The governance difference matters more than the code difference: Arbitrum upgrades pass a DAO vote and timelocks, OP Mainnet upgrades need the Security Council and the Foundation, and Base upgrades need Coinbase's multisig and its Security Council. Fees differ by about a cent per swap, so tooling, liquidity and governance weigh more. Arbitrum's delays give an app team more than two weeks to react to a pending upgrade; OP Mainnet and Base give none. Base's builders ship financial products on top of it: Coinbase launched tokenized stocks there in August 2026 with the B20 token standard (The Block, 2026).

Teams launching their own chain

Robinhood Chain runs on the Arbitrum stack, secures about $2.3B and sits at Stage 0 (L2BEAT, 2026). Unichain and Ink run the OP Stack and are Stage 1 (L2BEAT, 2026). Chains in the Superchain contribute the greater of 15% of net fee profit or 2.5% of gross fees to Optimism, in exchange for shared upgrades and a shared security process (Optimism docs, 2026). OP Mainnet itself contributes all of its revenue. Upgrade 20's reviewed proof setup covers OP, Ink, Soneium and Unichain; another OP Stack chain must build and verify its own (Optimism docs, 2026). Base chose a third route and now operates its own software. The choice is between joining a shared upgrade process, running Nitro under the Arbitrum DAO, or owning the stack and its maintenance.

Data current as of November 2026.

What Is Still Unsettled for Each Chain?

Each chain carries one open item that no source has dated: ZK proofs for OP Mainnet, a network token for Base and, for Arbitrum One, a pending L2BEAT review after its ordering change. None has a confirmed date as of 5 November 2026 (Optimism, 2026; Base docs, 2026; L2BEAT, 2026).

OP Mainnet and ZK proofs

Optimism named Succinct its first preferred ZK proving provider on 12 February 2026, with OP Mainnet first in line, and said validity proofs would remove the challenge period (Optimism, 2026). On 5 November 2026 L2BEAT still lists OP Mainnet as "Fraud proofs (INT)", and Optimism's challenger notes say ZK dispute games are not yet live in production (L2BEAT, 2026; Optimism docs, 2026). Upgrade 20 shipped Super Root games instead, and a later hard fork must still enable cross-chain messages. Succinct's own announcement said OP Mainnet would integrate OP Succinct "in the coming months", and nearly nine months on L2BEAT's label has not changed (Succinct, 2026). Until ZK proofs arrive, OP Mainnet keeps the seven-day withdrawal wait, and withdrawal tooling needs viem 2.51.0 or later.

Base on its own stack and the token question

Base reached Stage 1 on its own stack, and Cobalt, its third upgrade, went live on 30 September 2026, and Denim, which brings native 200-millisecond blocks, activated on 5 November 2026 (Base docs, 2026). Base's own blog says the chain is "moving away from the OP Stack", while L2BEAT keeps the OP Stack tag (Base, 2026; L2BEAT, 2026). Both labels are accurate for different things: lineage and specification on one side, the running software on the other. Beryl cut the single-proof withdrawal wait from seven to five days on 25 June 2026 (Base docs, 2026). A network token remains an exploration, and no source gives a date. Base's lead in value and traffic therefore comes without a token vote in its governance.

Arbitrum after Timeboost

L2BEAT flags Arbitrum One's page with a notice that impactful changes are pending and some information may be outdated (L2BEAT, 2026). The stage stays Stage 1 on 5 November 2026. Under the new rule a transaction with no tip can still be included, because waiting transactions earn a priority boost each round (Arbitrum Foundation, 2026). Arbitrum's sequencer is still a single operator, now ordering by priority fee, and the Security Council still holds an instant emergency path. A pending constitutional proposal would add an Ethereum-based voting path as a last resort if the DAO's L2 governance became unavailable (Arbitrum Foundation, 2026). Readers should check the L2BEAT page again before relying on any Arbitrum stage or delay figure quoted here, because the review is open.

FAQ

?Why do Base's docs give two withdrawal times, five days and one day?

Base's official bridge finalizes a withdrawal after a window that depends on how many proof types back the claim. A single proof needs five days since the Beryl upgrade of 25 June 2026. When both a trusted-execution-environment (TEE) proof and a zero-knowledge (ZK) proof back the same proposal, the window is one day. A challenge backed by a ZK proof stretches resolution to seven days (Base docs, 2026). Third-party bridges that use their own liquidity release funds sooner.

?Do users pay fees in ARB or OP, and does the token price change the cost of a transaction?

No. Users pay gas in ETH on Arbitrum One, OP Mainnet and Base (L2BEAT, 2026). ARB and OP are governance tokens, so a fall in either price leaves transaction costs unchanged. The tokens matter for votes and treasuries: ARB holders vote in the Arbitrum DAO, and OP holders vote in the Token House while a buyback program uses 50% of Superchain revenue (Optimism governance forum, 2026).

?Can the same Solidity contract run on all three chains without changes?

Standard Solidity contracts deploy on each, because L2BEAT lists the Ethereum Virtual Machine (EVM) as the virtual machine on all three (L2BEAT, 2026). Base keeps all existing remote procedure call (RPC) methods, including the optimism namespace, after its stack change (Base, 2026). Differences show up in fees, block times and chain-specific features such as Arbitrum One's WebAssembly virtual machine, so a test deployment on each chain's testnet comes first.

?Is a Stage 1 rollup as secure as Ethereum itself?

No. Stage 1 means a proof system works, the data sits on Ethereum and only a 75% Security Council can block or forge withdrawals (L2BEAT, 2025). It does not remove the single centralized sequencer, and L2BEAT states that stages measure decentralization, not protection from bugs. A Security Council can also upgrade contracts instantly.

?How does Robinhood Chain relate to Arbitrum?

Robinhood Chain is a separate chain that runs on the Arbitrum stack. L2BEAT lists it with about $2.3B of value secured and at Stage 0, below Arbitrum One's Stage 1 (L2BEAT, 2026). Arbitrum One is the chain governed by the Arbitrum DAO and covered by the figures in this article.

?What is the quickest way to move assets between Arbitrum, OP Mainnet and Base?

Official bridges send funds through Ethereum and wait out each chain's dispute window, so the quick route is a third-party bridge that fronts its own liquidity (Base docs, 2026). Each such bridge adds its own risk. L2BEAT tags the externally bridged share of value as resting on additional trust assumptions: 39.5% on Arbitrum One, 34.4% on Base and 19.2% on OP Mainnet on 5 November 2026 (L2BEAT, 2026).

?Where can a reader check a chain's current stage and value secured?

L2BEAT keeps one page per chain with its stage, state-validation label, sequencer-failure delay, exit window and value secured, and its figures change as upgrades land (L2BEAT, 2026). growthepie publishes fee tables by chain and hour. CoinPaprika lists ARB and OP prices and drawdowns. The figures in this article date from 5 November 2026, so a reader should reload those pages before acting.

?Does a larger value secured make a chain safer?

No. Value secured counts bridged and minted tokens and says nothing about governance or proofs. On 5 November 2026 Base held $14.8B and OP Mainnet $1.6B, yet both are Stage 1 (L2BEAT, 2026). On 5 November L2BEAT counted 39.5% of Arbitrum One's value as externally bridged, and about $1.14B of it as ARB itself. A chain's safety depends on its proofs, its upgrade rules and its sequencer, not on the size of its balance.

References / Sources