Why Does the Uniswap DEX Have Different Versions?

Open a token's pool list on Uniswap and you will often see the same pair several times, labeled v2, v3 or v4. Each label is a separate protocol with its own smart contracts, pools and liquidity. All four generations, from v1 in 2018 to v4 in 2025, still run at the same time.

This guide explains what each version changed, why the old ones are never switched off, how liquidity and volume split between them in October 2026, and what that means for traders, LPs and bulk trades on ChainTx.Tools.

#The short answer

Uniswap contracts can't be upgraded. So instead of patching the protocol, the team ships a new one next to it, and liquidity providers (LPs) decide whether to move. Each version fixed a specific limit of the one before.

*DefiLlama: TVL on October 11, 2026; volume from September 12 to October 11, 2026.

#Uniswap v1: the proof of concept

Hayden Adams deployed the first Uniswap contracts to Ethereum on November 2, 2018, the last day of Devcon 4 (project history). The design was minimal. A factory created one exchange contract per token, always paired with ETH. Prices followed the formula x·y=k, so the bigger the trade relative to the pool, the more the price moved. Every swap paid 0.3% to liquidity providers. The weak spot was token-to-token trading: swapping USDC for DAI went through two pools, USDC→ETH and ETH→DAI, which meant two fees and two price impacts.

#Uniswap v2: any pair, oracles and flash swaps

v2 went live on May 18, 2020 (launch post). The v2 whitepaper lists the main changes:

• any ERC-20/ERC-20 pair, so tokens can be paired directly;
• a price oracle: other contracts can read a time-weighted average price (TWAP) that is hard to manipulate within one block;
• flash swaps: take tokens from a pool and pay or return them by the end of the same transaction;
• an optional 0.05% protocol fee out of the 0.30%, which governance could turn on later.

v2 liquidity still covers the whole price range, from zero to infinity. That keeps it simple: deposit two tokens, receive LP tokens, and fees are added to the pool automatically. The downside is that most of that capital sits at prices the market never reaches.

#Uniswap v3: concentrated liquidity

v3 launched on May 5, 2021 (launch post). Each LP now picks a price range, and their capital only works inside it. The docs give an example: the v2 DAI/USDC pool used only about 0.5% of its capital for trades between $0.99 and $1.01, where almost all the volume happened (docs). In v3, an LP can put everything into that band. The announcement claimed up to 4,000× better capital efficiency than v2.

v3 also introduced fee tiers of 0.05%, 0.30% and 1%. A 0.01% tier for stablecoins was added by governance in December 2021 (proposal 9). Positions became NFTs, because every range is unique. Fees stopped compounding into the pool and now have to be collected.

The cost is complexity. If the price leaves your range, the position becomes 100% one token and stops earning fees until the price returns. So v3 liquidity needs managing, either by hand or through a vault.

#Uniswap v4: one contract, hooks and native ETH

v4 went live on January 31, 2025. The Uniswap Labs blog lists 10 networks at launch; the Uniswap Foundation lists 12 (Labs, Foundation). The liquidity math is the same as v3. What changed is the architecture around it (v4 vs v3):

• Singleton: all pools live in one PoolManager contract. Labs says creating a pool became up to 99.99% cheaper.
• Flash accounting: only net balance changes are settled at the end, so a multi-hop route moves real tokens just twice.
• Native ETH: no wrapping into WETH.
• Flexible fees: any fee from 0% to 100%, or dynamic fees that change from trade to trade.
• Hooks: external contracts that a pool calls before or after swaps and liquidity changes.

Before launch, v4 went through nine audits, a $2.35M security competition and a bug bounty of up to $15.5M. The docs now call it the recommended version for new integrations (overview).

#Why old versions keep running

The contracts are immutable. Core Uniswap contracts can't be upgraded, and nobody can pause them or reverse trades (docs). No admin key can move money from v2 pools into v4. When v2 was announced, the team wrote that v1 would keep working "for as long as Ethereum exists" (announcement).

Liquidity moves only when LPs move it. Each launch came with an optional migration tool, but many LPs had no reason to use it. v2 needs no attention, many older tokens keep their main pairs there, and other protocols are built around specific versions.

Routing hides the versions. Since December 2021 the Uniswap Auto Router has searched v2 and v3 pools together. It splits trades when that helps and adds a hop only if the result is better after gas (Auto Router v2). The Universal Router, launched in November 2022 together with the Permit2 approval contract, now combines v2, v3 and v4 swaps in one transaction (docs). UniswapX, launched in July 2023, adds signed orders. Competing "fillers" execute them and pay the gas, and a failed order costs the user nothing (docs).

#Forks and licenses

• v2 is under GPL-3.0, so anyone could copy it. SushiSwap (August 2020) and PancakeSwap (September 2020) started as v2-style AMMs, and DefiLlama lists hundreds of v2 forks.
• v3 Core was under the Business Source License 1.1, which barred commercial copies until April 1, 2023 (LICENSE). PancakeSwap v3 appeared a few days later (The Defiant).
• v4 Core is also BUSL-1.1 and becomes MIT no later than June 15, 2027 (LICENSE).

That's why so many DEXes on BNB Chain, Polygon or Base feel familiar: underneath they use Uniswap's v2 and v3 router interfaces.

#Uniswap in numbers (October 2026)

DefiLlama data as of October 11, 2026:

• v2 holds a quarter of the liquidity but less than 2% of the trading. It's passive capital that nobody has a reason to move.
• v3 and v4 split the volume almost evenly. On Ethereum, v4 already takes 58%; on Arbitrum and Base, v3 still leads.
• By network, Robinhood Chain (live since July 1, 2026) produced 40% of Uniswap volume over those 30 days, Ethereum 32% and Base 10%. Uniswap's share of all DEX volume was about 27%.
• Cumulative volume. Uniswap Labs announced passing $4 trillion in early December 2025 (Phemex News). DefiLlama's count is lower, so say whose figure you quote.

Fee switch. The "UNIfication" proposal passed in December 2025 with 99.9% support. It burned 100 million UNI and turned on protocol fees (proposal 93, Labs blog).

• December 2025: v2 pools and the largest v3 pools on Ethereum.
• March 2026: all v3 pools on Ethereum, plus v2 and v3 on eight more networks.
• July 2026: selected types of v4 pools on seven networks (proposal 100).
• Early October 2026: fees were running on Ethereum and eleven other networks, and Arc was added (proposal 102).

Collected fees can only be claimed by burning UNI.

#What the versions mean for traders and LPs

Fees. Where the fee switch is on, a v2 pool's 0.30% is split into 0.25% for LPs and 0.05% for the protocol. In v3, the protocol takes a quarter of the fee in the 0.01% and 0.05% tiers and a sixth in the 0.30% and 1% tiers (fee docs). The protocol share comes out of the LPs' cut, so traders pay the same total. In v4, though, a hook can add its own fee on top of the pool's.

Slippage. Concentrated liquidity makes pools deeper near the current price, but in a sharp move positions go out of range and that depth vanishes. A wide slippage tolerance keeps a trade from failing, but lets it fill at a much worse price.

Impermanent loss. In a v2 pool, if one token doubles against the other, the position ends up about 5.7% behind just holding both tokens (docs). A narrow v3/v4 range earns more fees per dollar but loses more when the price moves. A 2021 study of 17 v3 pools found that LPs earned $199.3M in fees and lost $260.1M to impermanent loss (Loesch et al.). A rough guide:

• v2: passive full-range liquidity, especially for volatile and tax tokens;
• v3: active ranges on liquid pairs, with the widest integrations;
• v4: like v3 without a hook, or with a hook whose code you trust.

Gas. A single swap costs about the same in every version. In Uniswap's own gas tests, a one-hop swap uses roughly 105,000–136,000 gas in v2, v3 or v4 (test snapshots). v4 saves on pool creation, multi-hop routes and native-ETH pairs, while a hook adds gas of its own.

#How ChainTx.Tools works with Uniswap versions

The EVM tools on ChainTx.Tools — bulk buy, bulk sell, multi-token buy and sell, and the EVM market maker — let you choose the DEX and version yourself. They don't aggregate across versions; every wallet trades through the router you select:

• Uniswap V2: the official Router02 with fee-on-transfer swap functions, so tokens with a transfer tax work.
• Uniswap V3: SwapRouter02. Before each trade the QuoterV2 contract checks the 0.01%, 0.05%, 0.3% and 1% pools and picks the best quote.
• Uniswap V4: the Universal Router, using only pools without hooks. Buying with the native coin needs no approval. The first sell of a token from a wallet needs a two-step approval (token → Permit2 → Universal Router); later sells don't.

The minimum output is always the quote minus your slippage. The versions on offer depend on the network: Base and Unichain, for example, have all three. The DEX list also includes Uniswap forks such as PancakeSwap, SushiSwap and QuickSwap.

#Summary

Uniswap has four versions because a protocol that can't be changed can only be improved by launching a new one next to it. v1 proved an on-chain AMM could work. v2 allowed any pair. v3 concentrated liquidity, which made it far more efficient but turned LPing into active work. v4 put every pool into one contract and made pools programmable through hooks. Old versions stay because their liquidity stays, and routers use all of them. In October 2026, v3 and v4 split the trading almost evenly, while v2 still holds a quarter of the liquidity. For a trader the version is mostly a routing detail. For an LP or a token team it is a choice between simplicity, efficiency and flexibility.