Layer 1 vs Layer 2 Blockchains: What's the Difference?

Introduction

Ethereum mainnet processed about 21 transactions per second in late October 2026, while the scaling chains built around it handled roughly 110 times that. Ethereum-ecosystem activity has moved to layer 2s (L2s), yet the guarantees behind those chains vary widely. Of the 83 layer 2s that L2BEAT, renamed ETHBEAT on 4 November 2026, tracked on 5 November 2026, ten had passed its Stage 1 or Stage 2 tests, and the Stage 2 group secured under $1M (L2BEAT, 2026). The layer 1 (L1) vs layer 2 question therefore matters more than the speed comparison usually attached to it. This guide explains what each layer does, how rollups differ from sidechains and validiums, and which checks show whether an L2 relies on its L1 for security or only borrows the name.

What Is the Difference Between a Layer 1 and a Layer 2 Blockchain?

A layer 1 blockchain orders, executes, settles and stores its own transactions with its own validators. A layer 2 blockchain executes transactions elsewhere and relies on a layer 1 to settle them and keep their data. In the Layer 1 vs Layer 2 split, "layer" describes dependency, not speed or quality.

What a layer 1 does

Bitcoin, Ethereum, Solana and BNB Chain ↗ are layer 1 (L1) blockchains. Each one runs its own consensus, in which miners or validators agree on the order of transactions, execute them and write the result into blocks that every full node can check against the shared ledger ↗. Security comes from the cost of attacking that validator set, and the native coin pays for it through fees and issuance. The trade-off is capacity. Every full node re-executes every transaction, so throughput stops at what ordinary node hardware can follow. Ethereum mainnet averaged about 21 transactions per second over the seven days to 4 November 2026, according to L2BEAT's activity tracker (as of 4 November 2026).

What a layer 2 does

A layer 2 (L2) is a separate network that executes transactions off the main chain and posts the results back to it. Arbitrum One, Base and OP Mainnet ↗ on Ethereum are the best-known examples, alongside the Lightning Network on Bitcoin. The L2 operator batches hundreds of transactions, compresses them and publishes the batch with a claim about the new state. The L1 stores that data and checks the claim. A user's funds therefore rest on L1 rules rather than on the operator's honesty. The payoff is volume: scaling projects tracked by L2BEAT averaged about 2,310 transactions per second in the same week, roughly 110 times Ethereum's own rate (as of 4 November 2026).

Data current as of November 2026.

That division of labour rests on a blockchain doing four separate jobs, and an L2 takes over two of them.

Which Jobs Does a Blockchain Do, and Which Can a Layer 2 Take Over?

A blockchain does four jobs: executing transactions, ordering them, settling disputes over the result and keeping the data available. An L2 takes over execution and ordering, then hands settlement and data back to the L1. That hand-back is the entire basis of an L2's claim to share the L1's security.

The four jobs of a blockchain

Execution means running transactions and computing new balances and contract states. Ordering, the consensus job, decides which transactions land and in what sequence. Settlement is the final court: the layer whose verdict on the state cannot be reversed without attacking the chain itself. Data availability (DA) guarantees that the raw transaction data was published, so anyone can rebuild the state and check it. A monolithic L1 such as Solana ↗ performs all four jobs on one validator set. Ethereum with rollups splits them, with rollups executing while Ethereum settles and stores the data. Modular blockchain designs ↗ go further and move the data job to a separate network such as Celestia or EigenDA.

What a layer 2 hands back to layer 1

An L2's sequencer orders user transactions and returns a near-instant confirmation, which is a promise rather than settlement. The promise becomes final once the batch reaches Ethereum. Since the Dencun upgrade in March 2024, rollups post that data as blobs ↗, a cheaper data format Ethereum keeps for about 18 days. That window is long enough for anyone to download and verify the batch. Alongside the data, the rollup posts a state root, a fingerprint of every balance after the batch. A zero-knowledge rollup (ZK-rollup) ↗ attaches a validity proof that the L1 contract checks immediately. An optimistic rollup lets the root stand unless someone proves fraud within a challenge window. Withdrawals to L1 become claimable after that step.

Rollups answer limited L1 capacity from the outside; the other answer is to enlarge the L1 itself.

How Do Layer 1 Blockchains Scale Without Relying on a Layer 2?

Layer 1 blockchains scale by raising the capacity of the base layer itself: bigger or faster blocks, quicker consensus or cheaper data. Every increase makes nodes harder to run, the tension the blockchain trilemma ↗ describes. Ethereum is now pushing its own L1 capacity after years of leaving growth to rollups.

Bigger blocks and faster consensus

Solana is the clearest case of scaling the base layer. It runs one chain with short block times, parallel transaction execution and a validator set built on high-end hardware. Sui and Aptos follow a similar monolithic path ↗ with their own parallel execution engines. The cost lands on node operators. Solana's validator hardware requirements sit far above what Bitcoin or Ethereum ask of a home node, which narrows who can verify the chain independently. Bitcoin chose the opposite end: small blocks that a laptop can follow, with payment scaling pushed to the Lightning Network. Each choice fixes a different point between throughput and the number of people able to check the ledger themselves.

How Ethereum is scaling its own layer 1

Ethereum adopted a rollup-centric roadmap ↗ in October 2020 and kept L1 changes conservative for years afterwards. The Fusaka upgrade went live on 3 December 2025 with PeerDAS, which lets nodes verify blob data by sampling pieces instead of downloading every blob (Ethereum Foundation, 2025). Two Blob Parameter Only forks then lifted the per-block blob target from 6 to 10 on 9 December 2025 and to 14 on 7 January 2026 (Ethereum Foundation, 2025). Fusaka also set the default gas limit at 60M, up from 30M at the start of 2025. Glamsterdam, which adds enshrined proposer-builder separation and block-level access lists, activated on the Sepolia testnet on 6 October 2026, and no mainnet date had been set (Ethereum Foundation, 2026). Its next testnet fork, on Hoodi on 26 October, was postponed on 22 October so developers can investigate how EIP-8037, a gas-cost increase for creating new state, interacts with existing mainnet contracts, and no new Hoodi or mainnet date had been set by 5 November 2026 (Ethereum Foundation, 2026).

Base-layer capacity sets the ceiling, and the label "layer 2" covers several designs that lean on that ceiling in very different ways.

Rollups, Sidechains or Validiums: Which Ones Count as a Real Layer 2?

By the definition ethereum.org and L2BEAT apply, a design earns the name layer 2 when users can rely on the L1 to rebuild the chain and force their funds out. Rollups meet that test because they publish data to the L1 and prove their state. Validiums and optimiums keep data elsewhere, and sidechains run on their own validators.

Rollups, optimistic and ZK

A rollup posts every transaction batch to its L1 and lets an L1 contract judge the resulting state. Optimistic rollups, including Arbitrum One, Base and OP Mainnet, treat a posted state as correct unless someone submits a fraud proof during a challenge window. That window keeps native withdrawals waiting about a week on Arbitrum One and OP Mainnet. ZK-rollups such as Starknet, ZKsync Era and Consensys' Linea network ↗ prove every batch with a validity proof before the L1 accepts it, so withdrawals skip the challenge period. The two families are converging. Base now pairs its fraud proofs with zero-knowledge components, and L2BEAT labels its system "Fraud proofs (1R, ZK)".

Sidechains

A sidechain is an independent blockchain connected to another chain by a bridge. It has its own validators, consensus and data, so the larger chain cannot rebuild its state or overturn a bad block. Polygon PoS is the best-known case. Marketing called it an Ethereum layer 2 for years, yet L2BEAT files it under "Others" with no stage and lists its data availability as "PoS network", which refers to Polygon's own validators (as of 5 November 2026). Sidechains are fast and cheap because they skip the L1 for security. A dishonest validator majority can freeze or redirect bridged funds, so a sidechain is exactly as secure as its validator set.

Validiums and optimiums

Validiums use validity proofs like ZK-rollups but keep transaction data off Ethereum, with a data availability committee or an external network. Optimiums do the same with fraud proofs. Both also add a trust assumption: if the external data holders withhold the data, users cannot rebuild the state to prove what they own. The Ethereum Foundation's March 2026 post describes such chains as inheriting "a subset of the security properties" of Ethereum. L2BEAT counted two validiums and one optimium, Celo, among the chains it tracks (as of 5 November 2026). Newer chains such as Plume's real-world-asset network ↗ post data to external committees and sit in its "Others" group.

Data current as of November 2026.

Even a true rollup can be controlled by a handful of key holders, which is the gap L2BEAT's stages were built to measure.

How Do L2BEAT Stages Show Whether a Layer 2 Inherits Ethereum's Security?

L2BEAT's stages measure how far a layer 2 has removed the power of a small group to override its code. On 5 November 2026 most value sat in Stage 1 rollups and almost none in Stage 2. For most L2s, the honest answer to "does it inherit Ethereum's security?" is partly (L2BEAT, 2026).

Stage 0, 1 and 2

L2BEAT introduced the framework in June 2023 from a 2022 proposal by Vitalik Buterin. Stage 0 is the entry bar: the project calls itself a rollup, posts state roots and data to Ethereum, ships source-available node software and runs a working proof system with at least five outside challengers. Stage 1 limits the override power. Only a compromise of at least 75% of the Security Council, or a bug, can block or forge withdrawals, and other upgrades must give users at least seven days to leave (L2BEAT, 2025). Stage 2 removes the training wheels: proofs are permissionless, users get at least 30 days before unwanted upgrades, and the council may act only on bugs provable on chain (L2BEAT, 2025).

What the November 2026 numbers show

L2BEAT tracked 83 active layer 2s securing about $38.8B (as of 5 November 2026); counting its 12 layer 3 projects, such as Hyperliquid, the total was 95 projects and about $45.4B. Only six were Stage 1: Base Chain, Arbitrum One, OP Mainnet, Starknet, Ink and Unichain. Four projects reached Stage 2 (Facet, Aztec Network, Ethscriptions and a Cartesi test deployment), and together they secured under $1M (L2BEAT, 2026). Stage 1 rollups held about $26.8B, roughly 69% of all value secured, and Base alone held about $14.8B. "Value secured" is L2BEAT's measure of every token bridged to or minted on a chain. It therefore runs higher than the decentralized finance (DeFi) deposit totals that DefiLlama reports as total value locked (TVL).

What a stage does not measure

A stage rates decentralization, not code quality. L2BEAT states this directly: a team can deploy an unsafe Stage 2 system with an experimental proof system or unaudited contracts. Stages also ignore the sequencer, the single operator that orders transactions on nearly every L2. They say nothing about fees, throughput or the bridge a user picks. Mantle shows the gap between labels, because L2BEAT lists it as a ZK-rollup with validity proofs yet still rates it Stage 0 (as of 5 November 2026). A Stage 1 badge answers one question, whether a small group can take user funds, and leaves the rest to separate checks.

Data current as of November 2026.

Table reflects L2BEAT data as of 5 November 2026.

Stage 1 status protects withdrawals, but daily use of an L2 still carries trade-offs that no badge records.

What Does Using a Layer 2 Cost in Trust, Speed and Convenience?

Using an L2 means trading some of the L1's guarantees for lower fees and faster confirmations. The costs come in three forms: trust in a sequencer and upgrade keys, waiting time on the way back to L1, and liquidity split across dozens of chains.

Sequencers, upgrade keys and exits

Nearly every L2 on L2BEAT's list runs one sequencer, operated by the team or company behind it (L2BEAT, 2026). That operator can delay or reorder transactions and, if it goes offline, stall the chain. The defence is forced inclusion: a user submits a transaction directly to the L1 contract, and the L2 must process it. On Base, L2BEAT records a maximum delay of 12 hours for that route (as of 5 November 2026). Upgrade keys are the second dependency, because whoever controls the contracts can change the rules. Fees are the main reason people accept these costs, and they depend on L1 data prices as much as on L2 traffic, as the breakdown of Ethereum gas fees and layer 2 savings ↗ shows.

Withdrawal delays, bridges and split liquidity

Deposits into an L2 arrive within minutes, but withdrawals through an optimistic rollup's native bridge wait out the challenge window ↗. Third-party bridges and exchanges shorten the wait by fronting the funds, and they add their own contract and custody risk. Bridges have been among crypto's most exploited infrastructure: the Ronin and Wormhole hacks of 2022 together took more than $900M. Liquidity also fragments. A token pool on Base cannot trade directly against the same token on Arbitrum One, so prices drift apart and large trades slip more than on one shared chain. Wallets and exchanges increasingly route across chains automatically, which hides the split without removing it.

Ethereum rollups are one scaling model among several, and the alternatives skip the L2 question entirely.

Where Do Solana, Avalanche L1s, Cosmos and Polkadot Fit in L1 vs L2?

Many scaling paths bypass Ethereum layer 2s altogether. Monolithic L1s such as TON, the chain tied to Telegram ↗, scale the base chain itself, while appchain frameworks give each application its own L1 that shares tooling or security with a wider network. Both models sit outside the L1 vs L2 blockchain split as Ethereum defines it.

Monolithic alt-L1s

Solana, Sui, Aptos, TON, BNB Chain and NEAR run execution, ordering, settlement and data on their own validators. Their pitch is one fast, shared state. Every app sits on the same chain, so a trade can touch several protocols in one transaction without a bridge. Sui's object-based design ↗ and Aptos and its Move language ↗ both come from engineers who built Meta's Diem project, and both execute transactions in parallel. NEAR splits its chain into shards ↗ that the protocol coordinates internally. The cost is the trade-off covered earlier: these chains ask more of their validators, and their own tokens carry the full security budget.

Appchain frameworks

Appchain frameworks let a project launch its own L1 while borrowing tools and, in some cases, security. Avalanche subnets became Avalanche L1s ↗ with the Etna upgrade, which dropped the rule that their validators must stake 2,000 AVAX on the main network. Cosmos chains share the Cosmos SDK ↗ and exchange messages over the Inter-Blockchain Communication (IBC) protocol, but each secures itself. Polkadot takes the opposite route. Its relay chain validates every connected chain, and since parachain auctions stopped on 19 September 2024, teams buy blockspace through Agile Coretime ↗. Each of these is a layer 1 in the security sense, whatever name the marketing uses.

Those alternatives keep pressure on Ethereum's design, which changed course at both layers in 2026.

How Is the Relationship Between Layer 1 and Layer 2 Changing in 2026?

The L1–L2 relationship is being renegotiated in 2026. Ethereum is scaling its own base layer again and calling it the hub for settlement and DeFi. Base, the biggest L2 by value secured, has taken control of its own software, and Polygon's zkEVM rollup has shut down (Base, 2026).

Ethereum is scaling layer 1 again

The Ethereum Foundation's platform team set out the new division of labour on 23 March 2026. L1 is to be "a truly permissionless and maximally resilient global hub for settlement, shared state, liquidity, and DeFi" that will "scale by orders of magnitude" (Ethereum Foundation, 2026). L2s keep a role, but the post frames their value as differentiation, customization and control rather than raw capacity. That marks a shift from the 2020 roadmap, which treated rollups as Ethereum's main path to blockchain scaling. The Fusaka blob increases and the 60M gas limit have already landed, and Glamsterdam has run on the Sepolia testnet, though its next testnet fork was postponed in October. More activity can therefore stay on L1 at tolerable fees, while L2s compete on features rather than cost alone.

Layer 2s are going their own way

Base, the largest L2 by value secured, announced on 18 February 2026 that it would leave the OP Stack for a stack it runs itself ↗ (Base, 2026). The switch took effect with the Azul hard fork on 28 May 2026, which runs only on Base's own clients and adds trusted execution environment (TEE) and ZK proofs, and Base kept its Stage 1 rating (Base, 2026). Polygon's zkEVM Mainnet Beta went the other way: its sequencer shut down on 3 July 2026, and a claims interface stays open until 31 December 2027 for funds still on it (Polygon, 2026). Value is concentrating too. Base and Arbitrum One together held about 63% of all L2 value secured (as of 5 November 2026).

Purpose-built L2s and appchains for tokenized assets ↗ show the same pull toward chains tailored to one job.

How Should Users Choose Between a Layer 1 and a Layer 2?

Choosing between a layer 1 and a layer 2 comes down to what a transaction is worth and how much trust the user accepts. Large, infrequent transfers favour the L1. Frequent small transactions favour a Stage 1 or Stage 2 rollup that posts its data to Ethereum.

Five checks before bridging

Five checks map to the risk categories L2BEAT publishes for each project. The first is the stage on L2BEAT, since Stage 0 means a small group can still change the rules or freeze withdrawals. The second is where the data lives: anyone can rebuild a rollup that posts to Ethereum, while an external DA layer adds a trust assumption. The third is the exit window: the days between an announced upgrade and its activation. The fourth is the sequencer and its forced-inclusion route. The fifth is the bridge itself, because a chain's native bridge inherits its security while a third-party bridge adds its own. The same checks apply whether funds arrive by bridge or by a direct exchange withdrawal.

Matching the chain to the job

A treasury moving seven figures once a month gains little from saving a few cents in fees and gives up L1 settlement guarantees for nothing. A user making dozens of swaps, game moves or payments a week pays far less on a rollup and can accept a sequencer's soft confirmation. Developers face a similar split. An app that needs to compose with DeFi protocols on Ethereum mainnet has reasons to stay on L1, while an app needing custom fees, privacy or throughput fits an L2 or an appchain. Rankings of the largest L2s ↗ change month to month, so the stage and data checks matter more than any one week's leader.

Data current as of November 2026.

FAQ

?Is Polygon a layer 1 or a layer 2?

Polygon PoS, the chain most people mean by "Polygon", works like a sidechain. It has its own proof-of-stake validators and stores its own data, and L2BEAT lists it under "Others" rather than as a rollup. Polygon also ran a ZK-rollup, Polygon zkEVM, but its Mainnet Beta sequencer shut down on 3 July 2026. Funds still on that chain can be claimed through an interface open until 31 December 2027.

?What is a layer 3 blockchain?

A layer 3 is a chain that settles to a layer 2 instead of directly to Ethereum. Teams build them to get their own fee rules, gas token or throughput while staying inside an L2 ecosystem, such as Orbit chains settling on Arbitrum One. Each extra layer stacks another set of trust assumptions, so a layer 3 can be no safer than the L2 beneath it.

?Do I need a separate wallet to use a layer 2?

No. Ethereum layer 2s such as Base, Arbitrum One and OP Mainnet use the same address format as Ethereum, so one wallet works on all of them once each network is added. You do need a small amount of ETH on each L2 to pay its gas fees, because balances on Ethereum mainnet cannot pay for transactions on an L2.

?Is Bitcoin's Lightning Network a layer 2 in the same way Arbitrum is?

Both settle on a layer 1, but they work differently. Lightning uses payment channels: two parties lock bitcoin on chain, trade signed balance updates off chain and settle when the channel closes. The design suits payments rather than general smart contracts, and needs funds locked in channels in advance. Arbitrum is a rollup that runs any Ethereum contract and posts every batch of data to Ethereum.

?Do I need ARB or OP tokens to pay fees on Arbitrum or OP Mainnet?

No. Arbitrum One and OP Mainnet charge gas in ETH, and ARB and OP serve mainly as governance tokens. Holders should know their track record: ARB sat 93.2% and OP 97.7% below their all-time highs, according to CoinPaprika (as of 5 November 2026). A rollup's security does not depend on the price of its governance token.

?Can I withdraw crypto from an exchange straight to a layer 2?

Yes, on most large exchanges. Coinbase, Binance and Kraken list networks such as Base, Arbitrum One and OP Mainnet as withdrawal options. Choosing the network the receiving wallet expects matters most, because funds sent on the wrong network can be hard or impossible to recover. A direct exchange withdrawal also skips bridge risk and the L1 gas fee of bridging yourself.

?What happens to my funds if a layer 2 shuts down?

On a rollup that posts its data to Ethereum, users can rebuild their balances from L1 data and withdraw through the L1 contracts, even without the operator. Polygon zkEVM followed that path after its sequencer stopped on 3 July 2026, and its claims interface stays open until the end of 2027. On a sidechain or validium, recovery depends on the operators and data holders cooperating.

?What is a based rollup?

A based rollup lets Ethereum's own block proposers order its transactions instead of running a separate sequencer. That removes the single-operator risk and inherits Ethereum's resistance to censorship, at the cost of slower confirmations tied to Ethereum's 12-second slots. Taiko launched its mainnet as a based rollup in May 2024 (Taiko, 2024), and the design now anchors the debate over L2 sequencers.

References / Sources