StakeWise (SWISE): osETH liquid staking through vaults, and why Aave’s adviser wants it frozen

StakeWise is a liquid staking protocol on Ethereum: you deposit ETH, it is staked by validators, and you can hold osETH, a token that keeps collecting staking rewards while you use it elsewhere. (The name is shared with an unrelated UK betting brand; this page covers the StakeWise crypto protocol only.) This week it is in the news for one reason. On 10 November 2026 LlamaRisk, a risk adviser to the Aave DAO, proposed freezing osETH as collateral on Aave V3 Core, and StakeWise asked for the token to be voted on separately. Nothing has been voted, and no vote date has been set.

The stakes for StakeWise are large. At 19:32 UTC on 13 November 2026, Aave V3 Core held 122,273 osETH as collateral, 82.4% of the 148,452 osETH in existence, according to the two contracts’ own on-chain data. Most of it sits in leveraged loops that borrow ether against osETH. This page explains how StakeWise’s vaults and osETH work, who holds the admin keys, what happened when the protocol clawed back stolen tokens after the Balancer exploit of November 2025, what the SWISE token gives holders, and what the Aave proposal would and would not change.

What is StakeWise? Liquid staking through a marketplace of vaults

Most liquid staking protocols run one pool. Lido and Rocket Pool take deposits into a single system and issue one token against it. StakeWise V3, live on Ethereum since 28 November 2023 according to the project’s blog, works the other way round: anyone can open a staking Vault, pick the node operator, set a fee and accept deposits, and stakers choose among those Vaults on app.stakewise.io.

Each Vault is a separate upgradeable smart contract built from modules for deposits and exits, validator registration, fees, MEV rewards and osETH minting. The documentation lists several kinds: standard Vaults, Vaults that issue their own ERC-20 share token, private Vaults with a whitelist (used for KYC pools), Vaults with a blocklist, and MetaVaults. A MetaVault runs no validators of its own; it spreads deposits across up to 50 sub-vaults through a “Curator” contract, and the only Curator today divides them evenly. MetaVaults became open to everyone on 21 May 2026, after a governance vote.

Isolation is the point of the design. A slashing penalty or an operator failure in one Vault hits that Vault’s stakers and not the others. The other side of it is that each Vault’s quality depends on who runs it, and the core audits listed further down do not cover third-party Vault code.

Upgrades follow the same split. Only a Vault’s own Admin can upgrade it, never the DAO. The DAO’s role is to register approved new implementations in a registry, which must be the same Vault type and exactly one version higher, and an Admin may leave a Vault on an old version indefinitely.

What staking with StakeWise costs

The Vault operator sets its own fee. The documentation describes 3% to 5% as typical, within a 0% to 100% range, and limits increases to 20% of the current fee at a time, with three days between changes. The DAO takes nothing for using a Vault. It earns only on osETH: 5% of the rewards that accrue on minted osETH, paid in osETH (the controller contract’s fee was set at 500 basis points on 13 November 2026).

The default path in the app is simpler than the marketplace suggests. “Simple staking” sends ETH through a MetaVault into the Genesis Vault, StakeWise’s main Vault and currently that MetaVault’s only sub-vault, and mints osETH for the user automatically. Leaving takes a fixed 15-hour claim delay plus Ethereum’s own validator exit queue. If an operator does not free the ETH within 24 hours, the oracles can force validator exits, according to the staker risk page.

How osETH works: an overcollateralised liquid staking token

osETH is not a receipt for a pool share. It is a token that a staker mints against ETH staked in a Vault, a little like a loan against collateral. Its counterpart on Gnosis Chain is osGNO. Its documentation describes a repricing token: your balance does not change, but each osETH is worth more ETH over time. At 19:33 UTC on 13 November 2026 the controller contract held 160,759.5 ETH of assets against 148,452.2 osETH, an exchange rate of about 1.0829 ETH per osETH.

How much osETH a staker can mint depends on the Vault. In a standard Vault the limit is 90% of the staked ETH; if a position reaches 92%, a liquidator can repay osETH and take the collateral with a 1% bonus. Vaults approved by the DAO, whose operator posts a bond of 5 million SWISE, can mint up to 99.99%, and in those Vaults liquidations are switched off at contract level. The Genesis Vault is one of them: on 13 November it held 162,515.9 ETH, with a 99.99% loan-to-value setting and a 5% fee.

That 99.99% figure carries weight in the design. It means almost all of the ETH in the main Vault can be turned into osETH, and that the protection osETH holders rely on is the rate at which the token is allowed to grow, not a cushion of extra collateral. The exchange rate is driven by a reward-per-second value that only the oracle aggregator, called the Keeper, can set. The docs say the oracles align it with the yield of the most exposed staker’s Vault so that osETH never grows faster than the ETH behind it.

How osETH keeps its peg to ETH

osETH has a soft peg to its exchange rate. When it trades above the rate, anyone can stake, mint and sell. When it trades below, holders can now queue osETH for redemption at the protocol rate. Redemptions were enabled by SWIP-41, posted on the StakeWise forum on 26 August 2026, and the redeemer contract was redeployed under SWIP-42 (28 October 2026) with a small fee equal to the rewards earned while waiting in the queue. On 13 November the osToken configuration pointed to the SWIP-42 redeemer.

Redemption is not fully automatic. A StakeWise-run service works out which positions can be redeemed (osETH minted by stakers but held elsewhere), StakeWise posts that list on-chain as a Merkle root, and the oracles force validator exits if the queue stalls. On DefiLlama, osETH priced at $2,093.61 at 18:52 UTC on 13 November while ether priced at $1,940.13, a ratio of 1.0791, about 0.35% below the protocol rate.

Boost: the osETH loop on Aave

StakeWise’s own leverage product, Boost, is the main reason osETH and Aave are so closely tied. Available on Ethereum from 0.05 ETH, Boost supplies a user’s osETH to Aave, borrows ETH against it, stakes that ETH back into the Vault and repeats the loop automatically. The Boost documentation puts the result at about 6 times in a 90% Vault and up to 14 times in a Vault near 100%.

The trade earns staking rewards on the whole position and pays Aave’s ETH borrow rate on the debt. The docs give a historical total yield of “approximately 4-6%” and warn that it turns negative if borrowing costs stay above staking yields for long. That happened briefly between 20 and 25 July 2025, when Aave’s ETH borrow rate jumped; StakeWise then proposed on its forum to refund 5.38241 osETH of fees to Boost users.

Boost labels a position “Moderate” above 93.8% loan-to-value and “Risky” above 94.5%. At 94.5% anyone can trigger an unboost on the holder’s behalf, and the core team says it watches positions to do so; the docs call it a final safeguard. Aave prices osETH with StakeWise’s own exchange-rate feed rather than its market price, so a discount on exchanges does not by itself push Boost positions towards liquidation. StakeWise’s founder argues that its leverage contracts are immutable and return borrowed ETH through validator exits rather than swaps on exchanges.

Who controls StakeWise: a 4-of-7 Safe, 11 oracles and 2-of-2 vault admins

The protocol has no single admin key, but power is concentrated in a few places. The StakeWise DAO treasury is a Safe multisig with seven signers and a threshold of four, read on-chain on 13 November 2026. The treasury page names the signers as Ali Nuraldin, Andreas Dittrich, Auryn Macmillan, Dmitri Tsumak, Gleb Dudka, Mikko Ohtamaa and Otto de Voogd, and the seven owner addresses on-chain match that list.

That Safe owns the contracts that matter most: the Vaults registry, the osToken configuration (which sets each Vault’s loan-to-value, liquidation threshold and the redeemer), the osETH controller (fee and capacity), the Keeper (the oracle set) and the node manager. No timelock contract sits in front of them. The delay comes from the governance process instead: approved Snapshot votes are executed through the SafeSnap module after a bond-and-challenge window and a 24-hour cooldown, during which four of the seven signers can reject a transaction that the DAO did not approve.

Two other keys are smaller still. The osETH redeemer is owned by a 2-of-2 Safe, and so is the Genesis Vault, which backs most of the osETH in circulation. One signer of each is an address the docs list for Dmitri Tsumak; the other signers are not publicly identified. Because only a Vault’s Admin can upgrade it, the DAO cannot overrule that 2-of-2 on Genesis Vault upgrades.

The oracles are a committee of 11 per chain, and the Keeper needs 6 of the 11 signatures to update rewards or register validators. They hold no funds. On Ethereum the operators are Chorus One, Stake.fish, Telekom, Finoa Consensus Services, Bitfly, SenseiNode, Gateway.fm, the Gnosis Chain team, P2P, DSRV and StakeWise Labs, the core development team.

Governance votes with few voters

Changes go from a forum idea to a SWIP (StakeWise Improvement Proposal) with a five-day poll, then a seven-day Snapshot vote. Submitting a vote takes 1 million SWISE; a vote passes if yes beats no and at least 3 million SWISE take part. After a vote ends, someone must post a SWISE bond asserting the result, which opens a 24-hour challenge window. SWIP-40, voted on Snapshot from 6 to 11 May 2026, raised that bond from 200,000 to 1 million SWISE, because 200,000 SWISE had come to cost less than $1,000, its forum post said.

Turnout is thin. On Snapshot, the re-run of SWIP-39 in June 2026 drew five voters and 8.96 million SWISE, with the largest voter holding 50.5% of the votes cast. SWIP-40 drew three. Across the dozen Snapshot votes since September 2025, turnout ranged from 3 to 13 voters, so a small group of holders can pass proposals on their own.

The Balancer exploit of 3 November 2025 and the osETH clawback

StakeWise’s contracts have not been hacked, but its admin powers were tested once. On 3 November 2025 an attacker drained Balancer V2 composable stable pools, including the osETH-ETH pool on Ethereum and the osGNO-GNO pool on Gnosis Chain. A StakeWise forum post of 25 November 2025 counts about 6,851 osETH and 13,495 osGNO taken from those pools.

The same day the seven members of the DAO multisig used the DAO’s ownership of the token contracts. As DL News reported on 4 November 2025, they gave the DAO wallet temporary controller rights over osETH and osGNO, burned the tokens sitting in the attacker’s wallet, minted the same amounts to DAO wallets on Ethereum and Gnosis Chain, then revoked the rights. StakeWise recovered about 5,041 osETH, nearly three-quarters of what was taken and worth about $19 million according to DL News, and all 13,495 osGNO, about $1.7 million.

StakeWise told DL News that “Monday November 3 was the first time that the powers of the emergency multisig were invoked.” It added: “We believe it would be right to also make it the last such time.” DL News wrote that the recovery, welcome as it was for victims, exposed how centralised StakeWise was: a multisig had been able to burn and mint tokens in other people’s wallets.

Governance followed. SWIP-37, posted on the forum on 4 November 2025, proposed that the DAO renounce ownership of the osETH and osGNO token contracts so that no one could again appoint a controller able to mint or burn. Its Snapshot vote ran from 5 to 12 December 2025 and passed with 5.76 million SWISE in favour, none against, from 11 voters. On 13 November 2026 the osETH contract’s owner was the zero address. The proposal tied the osGNO step to the arrival of Boost on Gnosis Chain. A separate vote, also opened on 5 December 2025, approved a pro-rata distribution of the recovered tokens to the liquidity providers who lost them; it has not been confirmed that every claim has been paid.

Renouncing the token contracts did not remove all DAO power. SWIP-37 described the power left as approving new Vault versions, but the treasury Safe still owns the osToken configuration and the controller, which set loan-to-value limits, fees and the redeemer.

Why osETH is in Aave’s freeze proposal

LlamaRisk’s proposal, posted on the Aave governance forum on 10 November 2026 at 19:33 UTC, is an ARFC: a request for comment, the discussion stage before any vote. It covers rsETH, osETH, LBTC, tBTC and rETH on Aave V3 Core, Aave’s main Ethereum market, plus a few small reserves; rsETH and tBTC were already frozen. A freeze stops new supply, sets the asset’s loan-to-value to zero, including in every E-Mode it belongs to, and leaves existing positions open: holders can still repay, withdraw or migrate. Our news report on the Aave freeze proposal covers the full package.

LlamaRisk cites no hack or depeg. Its argument is about leverage and who earns from it. According to its figures as of 9 November, osETH backed $256.6 million of WETH debt with $279.4 million of collateral, at about 14 times leverage, and the ten largest positions held 66% of the debt. osETH already has a loan-to-value of zero outside Aave’s ETH-correlated E-Mode and 93% inside it, and the freeze would cut that 93% to zero, ending new borrowing against it. Because osETH shares that E-Mode with wstETH, weETH, cbETH and WETH, LlamaRisk writes, its exposure cannot be reduced through the category’s liquidation threshold. As a last step, if loops do not unwind, the proposal allows the WETH base borrow rate on Core to rise by up to 1 percentage point, in half-point steps at most monthly, with rebates for borrowers not using the affected collateral. Aave’s osETH supply cap is 123,000; on 13 November the reserve was 99.4% full.

StakeWise’s founder, signing as Dmitri, replied on 11 November at 14:02 UTC. He wrote that StakeWise has run liquid staking since January 2021 without a hack or a validator slashing, and that osETH has been Aave collateral for two and a half years. His request: “we respectfully ask LlamaRisk and the Aave DAO to move the proposed osETH freeze to a separate vote”. Unchained reported the proposal and the reply the same day.

By 19:35 UTC on 13 November LlamaRisk had not answered in the thread, no vote had been scheduled, and a forum timer was set to close the thread 30 days after its last post, on 13 December 2026. If the freeze passed, new borrowing against osETH on Core would stop, which would block new Boost loops there. StakeWise has not said what it would do with Boost in that case.

StakeWise TVL and fees on DefiLlama

DefiLlama counted $763.0 million in StakeWise at 18:28 UTC on 13 November 2026, almost all on Ethereum ($1.1 million on Gnosis Chain). That is down from $881.4 million on 7 November and well below the peak of $1.58 billion on 7 October 2025.

The fall of the past week comes from the price of ether, not from withdrawals. Measured in ETH, DefiLlama counts about 393,000 in StakeWise on 13 November against 392,743 on 7 November, slightly more. Lido and Rocket Pool lost a similar share of their dollar value over the same seven days.

DefiLlama counts as fees the staking rewards collected by the Vaults, and as revenue the DAO’s cut: about 5% on osETH and osGNO and about 10% on the older rETH2 token. Most of the value goes to stakers and operators.

SWISE token: governance votes and no revenue share

SWISE is StakeWise’s governance token, launched on 26 April 2021 and distributed in part to early stakers. Its total supply is 1 billion, according to the tokenomics page: 510 million for the community over four years, 251.5 million for investors (six-month cliff, then 18 months), 217 million for the team (six-month cliff, then 42 months) and 21.5 million kept for future fundraising. Under that published schedule, team and investor tokens have finished vesting.

Holding SWISE gives voting power in the DAO, and nothing else. DefiLlama’s methodology says there is no revenue share to SWISE holders. The treasury, which holds mainly osETH and SWISE, earned about 20 osETH a month in fees and held about 370 osETH, according to a forum post of 3 September 2025. A SWISE buyback of up to 35 osETH a month was proposed in that post; no vote on it has been found. DefiLlama put SWISE’s market value at about $2.2 million around 19:20 UTC on 13 November 2026, small next to the $763 million the protocol holds.

StakeWise audits

StakeWise publishes its audit reports in the v3-core repository on GitHub. Eight cover V3: Halborn (May and August 2023), Sigma Prime (August 2023, June 2024 and September 2024), Consensys Diligence (March 2024), ABDK (September 2025) and Statemind (March to April 2026).

The Statemind report covers the newest code: the MetaVault refactor, changes to the osETH redeemer and the node manager that lets operators join by posting a bond. It found 29 issues, none critical or high, three medium and 26 informational; 19 were fixed and 10 acknowledged. One acknowledged medium issue lets a depositor capture some sub-vault rewards through a stale MetaVault share price; the team argued the gain is limited to about 12 hours of rewards. No public bug bounty programme has been confirmed.

StakeWise vs Lido and Rocket Pool

On DefiLlama, at the same moment on 13 November, Lido held $18.9 billion, ether.fi $3.7 billion and Rocket Pool $939 million, against StakeWise’s $763 million, ninth among liquid staking protocols. The tokens differ in kind. stETH and rETH, explained in our guide to liquid staking tokens, are issued by one system against its pool of validators. osETH can be minted from any StakeWise Vault, carries the DAO’s 5% cut of rewards, and depends on the rate the oracles set rather than on a single pool’s balance.

For a staker the choice is between picking an operator yourself, and bearing that operator’s risk, or accepting the defaults of a larger single pool. For an osETH holder, much of the token’s market depends on Aave.

History, Gnosis Chain and what comes next

StakeWise’s first pool went public in March 2021 with a two-token design, sETH2 and rETH2. A $2 million seed round, led by Greenfield Capital with Collider Ventures, Gumi Cryptos and Lionschain Capital, is dated 8 March 2021 on DefiLlama, and Blockdaemon and boldstart ventures made a strategic investment of an undisclosed amount announced on 2 March 2022. V3 was announced on 28 September 2022 and went live on Ethereum on 28 November 2023.

MetaMask Pooled Staking launched on StakeWise Vaults in June 2024. V3 reached Gnosis Chain in July 2024, with osGNO, but Gnosis remains marginal at about $1.1 million. Obol distributed validators joined the Genesis Vault on 2 April 2025, and the older V2 pool and StakeWise Solo were deprecated from 15 April 2025.

StakeWise has published no dated roadmap beyond what is live: permissionless MetaVaults since May 2026, osETH redemptions since August to October 2026 and a node manager for bonded operators audited in spring 2026. The date that matters for osETH holders is outside StakeWise’s control: whether, and when, the Aave DAO votes on LlamaRisk’s proposal.

Risks to weigh before staking with StakeWise

• Most osETH sits on one market: 82.4% of supply was Aave V3 Core collateral on 13 November 2026, largely in loops of about 14 times.
• The Genesis Vault mints up to 99.99% of its stake as osETH with liquidations off, so osETH relies on the oracle-set exchange rate staying below the Vault’s real yield.
• Key powers sit with a 4-of-7 Safe without a timelock contract, and the Genesis Vault and osETH redeemer are each controlled by a 2-of-2 Safe.
• Snapshot votes since September 2025 drew between 3 and 13 voters.
• Leaving takes a 15-hour delay plus Ethereum’s exit queue, and redemptions depend on a list StakeWise itself publishes.

None of this is a sign of a problem today: there is no depeg, and StakeWise’s contracts have no recorded hack. The next fact to watch is LlamaRisk’s answer to StakeWise in the Aave thread, and whether osETH gets the separate vote its issuer asked for.