Solend Is Now Save Finance: What the On-Chain Data Says (2026)

Solend still exists. It is called Save now, it holds roughly $84.5 million of net deposits on Solana, and the lending program it launched in 2021 has never stopped running. What has stopped is almost everything written about it: the documentation that ranks first on Google for the word "solend" was last touched three to four years ago, it names an upgrade authority that no longer controls the program, and the page that was supposed to describe the protocol's flash loans still links to a pull request that was closed without ever being merged.

This article is the result of one afternoon spent checking the Solend brand against primary sources — the Solana mainnet RPC, DefiLlama, GitHub, the Save documentation itself, and the two websites still serving under two different names. Every number below carries the date and the endpoint it came from, because the whole point is that the published numbers disagree with each other.

Yes — but under a different name. On 24 July 2024 the team announced that Solend had been rebranded to Save, "Solana's permissionless savings account," and launched three new products alongside it: SUSD (a 0% interest stablecoin borrowed against SOL), saveSOL (a liquid staking token), and dumpy.fun (a platform for shorting memecoins). Decrypt covered the announcement the next day, and at the time read $395 million of deposits and $92.9 million of borrows off the app's own dashboard.

Two years and three months later, here is the state of the Solend/Save estate as of 17 October 2026, measured with plain HTTP requests:

Property	Status	What it serves
save.finance	200	The current app, on Vercel, last modified 15 Oct 2026
solend.fi	200	The old Solend app, still live on Netlify, 11.4 MB bundle
www.solend.fi	301	Redirects to save.finance
docs.solend.fi	400	Nothing to a plain client — and it is Google's top result
docs.save.finance	200	The Solend-era docs, retitled "Save (formerly Solend)"
app.solend.fi / app.save.finance	NXDOMAIN	Neither resolves; the app lives at the apex domain
dumpy.fun	200	Live, but not linked from the Save navigation

So the protocol is alive and actively deployed — the main site was rebuilt two days before this article was written. The problem is not abandonment. The problem is that the Solend brand has been left standing next to the Save brand, with a stale documentation set in between, and search engines still point newcomers at the older half.

It is worth separating that from a genuine shutdown. When we audited Step Finance, the company had publicly ceased operations and the staking app was still accepting wallet connections eight months later. Save is the opposite case: the operating company is still shipping, and the dead surface is documentation and a legacy domain rather than the product itself. Both are confusing for a user; only one is dangerous.

Solend was once the largest DeFi protocol on Solana, and it is now the fourth-largest lender on its own chain. DefiLlama has 1,886 daily data points for the protocol going back to 12 August 2021, which makes the arc easy to read without anyone's marketing attached.

Date	TVL (DefiLlama)	Context
12 Nov 2021	$928,478,641	All-time peak
1 Jul 2022	$224,346,001	Days after the whale governance vote
1 Jan 2023	$16,448,572	Post-FTX trough — 98.2% below peak
24 Jul 2024	$278,460,948	The day the Save rebrand was announced
24 Nov 2024	$493,893,479	Post-rebrand peak
1 Jan 2026	$124,984,127	—
17 Oct 2026	$84,547,901	Today, with $159,128,160 borrowed

Three percentages fall out of that series. Save is 90.9% below its November 2021 peak, 82.9% below its post-rebrand peak of November 2024, and 69.6% below where it stood on the day the rebrand was announced. The second of those is the one that matters, because it is measured entirely inside the Save era and so cannot be blamed on the 2022 bear market.

The borrow side is healthier than the headline suggests. DefiLlama reports $159,128,160 borrowed against that $84.5 million of net TVL, which implies roughly $243.7 million supplied and a blended utilisation around 65.3%. A lending market with two-thirds of its book lent out is not an empty building. It is a smaller one than it used to be.

The Eclipse deployment is a different story: DefiLlama carries it at $0. A second chain was opened and never filled, which is a pattern worth recognising on any multi-chain protocol — check each chain's TVL separately rather than trusting the chain list on a landing page.

Use save.finance. That is the maintained front end, and its HTTP headers show a Vercel deployment last modified on 15 October 2026 — two days before this was written. Its navigation reads Dashboard, Pools, Rewards, saveSOL, Revenue, Security, Docs.

But solend.fi also returns HTTP 200, from Netlify, and it is still the complete legacy Solend application: an 11.4 MB JavaScript bundle, Amplitude analytics keyed to the old project, Open Graph tags pointing at @solendprotocol, and a content-security-policy that still whitelists Squads v3, Sentre and Snowflake Safe as permitted frame ancestors — the multisig front ends a DAO treasury would embed it inside. The footer on the Save app, meanwhile, reads "Save 2025."

Two live apps under two brand names, pointing at one program, is a phishing surface whether or not anyone is currently abusing it. If you search the word "solend" today, the first result is docs.solend.fi, which returns HTTP 400 to an ordinary client, and the fifth is the legacy app. Nothing on that page tells you the protocol has a different name now.

Two practical rules follow. First, verify the domain before connecting a wallet, the same discipline we recommend for any Solana token or contract in our rug check walkthrough. Second, check the program your transaction is actually touching — the program ID below is public, and Solscan will show you which program a simulated transaction invokes before you sign it.

The Solend lending program is So1endDq2YkqhipRh3WViPa8hdiSpxWy6z3Z6tMCpAo. A getAccountInfo call against api.mainnet-beta.solana.com confirms it: owned by BPFLoaderUpgradeab1e11111111111111111111111, executable: true, with its program data in a 1,240,333-byte account at DMCvGv1fS5rMcAvEDPDDBawPqbDRSzJh2Bo6qXCmgJkR holding 8.63 SOL of rent.

That account also carries the upgrade authority and the slot of the last upgrade, and this is where the documentation falls apart. Three different addresses are presented as the upgrade authority, and none of the two documented ones is the live one.

Source	Address given as upgrade authority
Docs table, Access Controls and Addresses pages	2Fwvr3MKhHhqakgjjEWcpWZZabbRCetHjukHi1zfKxjk
Docs prose, one sentence later on the same page	GDmSxpPzLkfxxr6dHLNRnCoYVGzvgc41tozkrr4pHTjB
Solana mainnet, 17 Oct 2026	RY93CZYe5g6drtG7W9PmHRPzaBLZ1uwihTzayQTmJfh

The second of those addresses is listed on the project's own DAO page as "other treasury operations," not as an upgrade key. The third — the one that actually holds the authority — is an account that does not exist on chain at all: getAccountInfo returns null for it, meaning zero lamports and zero bytes of data. That is legal and unremarkable on Solana, because an upgrade authority only has to produce a signature, not hold a balance. It does mean you cannot learn anything about it from the chain.

What you can learn comes from the docs themselves, and it is unusually candid. The Access Controls page says, verbatim: "We hold the keypair of the Upgrade Authority, so we utilize that to push upgrades to mainnet." It continues: "For now, there isn't a timelock program, but transitioning to governance in the future will cause upgrades to go through a governance vote as well." That page is marked "Last updated 3 years ago." The future it describes has not arrived.

Two more checks are worth running on any Solana protocol holding deposits, and both return something here:

• When was the program last upgraded? The program data account records slot 350,614,570. Passing that to getBlockTime dates the upgrade to 2 July 2025, 10:58 UTC — fifteen months ago.
• Does the deployed binary match the public source? verify.osec.io/status/So1endDq2YkqhipRh3WViPa8hdiSpxWy6z3Z6tMCpAo returns is_verified: false with an on-chain hash of ef621872… against an executable hash of 4f3156c6…, and a last check dated 2 July 2025 — the same day as the upgrade. The docs predicted this too: "Our code is not 'Anchor verified' yet, but we will look into doing so soon."

None of that is evidence of wrongdoing. The source is genuinely public, the last code change on the mainnet branch of solendprotocol/solana-program-library is dated 2 July 2025 at 10:42 UTC — sixteen minutes before the on-chain upgrade, which is exactly what you would expect from an honest deploy. The point is narrower and still worth making: a protocol holding $243.7 million of supplied assets can be upgraded by one signature, with no timelock, and the three addresses published as the holder of that signature do not agree with each other or with the chain.

A note on GitHub, because it misleads here in a way worth knowing. The repository's pushed_at field reads 4 February 2026, which is what most "is this project alive?" badges display. The last commit on the default branch is 2 July 2025. pushed_at counts a push to any branch; always read commits?per_page=1 on the default branch instead. Eleven distinct authors appear in the last 100 commits, and the two repositories a developer would reach for first — solend-sdk (43 stars) and liquidator (190 stars) — have both been archived since October 2022.

The reason unilateral upgrade authority is not an abstract concern at Solend specifically is that the protocol once proposed using it to take control of a user's account, and the DAO said yes.

On 19 June 2022, co-founder Rooter posted governance proposal SLND1. The protocol's largest account, 3oSE9CtGMQeAdtkm2U3ENhEpkFMfvrckJMA8QwVsuRbE, had 5.7 million SOL deposited — around $170 million at the time — against $108 million borrowed in USDC and USDT. That single wallet represented 25% of total TVL, 95% of the Main Pool's SOL deposits and 88% of its USDC borrows, with a liquidation price of $22.30 and no on-chain activity for twelve days.

The proposal asked holders to enact special margin requirements for any borrower above 20% of the pool and to "grant emergency power to Solend Labs to temporarily take over the whale's account so the liquidation can be executed OTC." The mechanism was stated plainly in the proposal: "This would be done via a smart contract upgrade."

It passed. Voting was open for six hours. 97.5% of participating votes were in favour, and it cleared a 1% quorum with 1.13% of supply — a margin that, as CoinDesk and Fortune both reported at the time, rested on a single voter whose account was the only reason the threshold was met.

The backlash was immediate, and on 20 June SLND2 invalidated SLND1 and extended the governance voting window from six hours to one day. It carried 1,480,264 votes to 3,535 — 99.8% — and CoinDesk reported that one wallet paid roughly $700,000 for additional voting power, ending up with about 90% of the votes cast. A governance crisis resolved by buying a supermajority is not a reassuring resolution, but it did stop the takeover.

What happened to the whale? SOL recovered, the liquidation never triggered, and the account was never seized. Querying getSignaturesForAddress today, that wallet's last on-chain transaction was 18 March 2024, and it holds 1.62 SOL. The crisis resolved itself by not happening.

Four years later the specific capability that made SLND1 possible is unchanged: one keypair, no timelock, an unverified build. The docs page describing "Recovery Mode" still notes that during heightened volatility "a council of contributors can make parameter changes with greater flexibility," and that those changes "might lead to liquidations occurring, done so without liquidation penalty in market turmoil." It is an honest disclosure, and it is three years old.

The fee schedule is one of the few parts of the Solend documentation that is both specific and still broadly applicable, because fee parameters are set per reserve rather than hard-coded. As published:

• Interest rate spread: 20%. One-fifth of the borrow interest paid by borrowers is kept by the protocol rather than passed to suppliers. This is the main revenue line.
• Origination fee: 10 basis points (0.1%) on most vaults, charged once, upfront, when the loan is taken out. It splits 80/20 between the protocol and the "host" — any interface that refers the loan collects the 20%.
• Liquidation fee: 30% of the liquidation penalty, routed to the DAO treasury. The penalty itself is a per-reserve parameter, so the amount a liquidated borrower loses depends on which asset they posted.
• Solana network costs: 0.000005 SOL per transaction, plus roughly 0.01 SOL of rent the first time an account is created for you. The docs note the rent "can be recuperated later when the account is closed (though this isn't yet implemented on solend.fi)."

That last parenthetical is a small thing that tells you a lot: a feature described as not yet implemented, on a domain that has since been renamed, in a page last updated three years ago. If you want the live per-asset numbers you have to read them in the app, because the only place the origination fee, flash loan fee, host fee percentage and liquidation penalty are published per reserve is the interface itself.

One hard limit worth knowing before you plan a strategy: an account can hold at most six positions, where a position is any supplied or borrowed asset. The docs attribute this to Solana's compute budget and suggest multiple accounts as the workaround. That constraint is four years old and has survived several rounds of Solana fee and compute changes, so treat it as current until the app tells you otherwise.

Solend has a flash loan instruction, but the safe, composable version the documentation promises was never delivered. This is one of the most-searched questions about the protocol and the honest answer is three sentences long, so here it is in full.

The Flash Loans page says: "Currently, this version of flash loan implementation for Solend exists, but functionality is limited due to the reentracy of Solana transactions, where Program A calls Program B, and etc. We are currently working on a working flash loan program and the mitigation for it, so that it can be used safely, and will document it here when it's ready." That page is marked "Last updated 4 years ago."

It links to pull request #70 on solendprotocol/solana-program-library, titled "Composable flash loan implementation." Checking it via the GitHub API: opened 9 February 2022, eleven commits, five comments, last activity 25 January 2023, and merged_at: null — closed without being merged. The documentation still says the team is "currently working on" it and will document it "when it's ready."

If you are building a liquidator or an arbitrage bot against Solend, that is the state of play: the original limited instruction is in the deployed program, the improved one is not, and the reference liquidator repository has been archived since October 2022. For most on-chain strategies on Solana today, flash liquidity comes from elsewhere — and if what you are actually after is automated execution rather than atomic leverage, the honest comparison is against purpose-built tooling, which we cover in our guide to the best Solana trading bots.

DefiLlama's yields endpoint lists 59 Save pools summing to about $100.9 million. The distribution is the interesting part.

37 of the 59 pools (62.7%) pay 0% APY, and those pools hold $83.9 million — 83.1% of the total. That is not a failure; it is what the protocol is for. The majority of what sits in Save is liquid staking tokens posted as collateral, earning their staking yield from the LST itself rather than from Save, so there is no borrow demand against them and nothing to pay out.

Pool	TVL	Supply APY
mSOL	$25,651,599	0%
bSOL	$22,493,374	0%
JitoSOL	$8,943,438	0%
DAI	$7,984,370	0%
SOL	$7,656,785	1.87%
saveSOL	$5,155,369	0%
USDC	$4,007,408	3.82%
cbBTC	$2,250,343	0.41%
USDT	$1,627,784	2.91%

Three things to take from that table. mSOL and bSOL alone are 47.7% of everything in Save, which makes this a concentrated venue rather than a broad money market — if you want to understand the deposits, start with Marinade and the rest of the Solana liquid staking complex. SOL and SOL-derivative pools are $75.0 million across 17 pools, or 74.3% of the total, so Save in 2026 is overwhelmingly a leveraged-staking venue. And the headline APY is a trap: the highest-paying pool in the whole set is a USDC reserve advertising 16.46% on $16,209 of deposits — real, but too shallow to deploy into, and a reminder to read the TVL column next to every yield figure.

saveSOL, the LST launched with the rebrand, is tracked separately at $6.27 million and 4.91% APY. Of the three products announced in July 2024, it is the one that unambiguously shipped and is visible in the navigation. SUSD does not appear anywhere in the current documentation index or the site's navigation. dumpy.fun resolves and returns a page, but is not linked from Save.

Context matters more than the raw number. Here is where Save sits among Solana lenders, taken from DefiLlama's protocols endpoint on 17 October 2026:

Protocol	Chain	TVL	7-day change
Kamino Lend	Solana	$1,306,171,993	−1.5%
Jupiter Lend	Solana	$1,088,491,991	−4.2%
Suilend (same team)	Sui	$158,190,562	−7.5%
Save (formerly Solend)	Solana	$84,547,901	−4.9%
marginfi Lending	Solana	$35,998,975	−9.2%

Save holds 6.5% of Kamino's TVL. The protocol that was once the biggest thing in Solana DeFi is now a fifteenth the size of the current leader. Our write-ups of Kamino Finance, Jupiter Lend, marginfi and Loopscale cover what the rest of that table is doing differently.

The row that explains the most is the third one. Suilend, launched by the same team on Sui in March 2024, holds 1.87× Save's TVL. That is not an accusation of neglect — a team is allowed to build two things — but it is visible in more than the TVL. The DAO wallet table in Save's own documentation lists two Sui treasury operations addresses alongside the Solana ones, and the solendprotocol/sips repository on GitHub is titled "Sui Improvement Proposals." When the founder spoke to Decrypt at the rebrand, he described building on Ethereum and Solana as working "with chisels and hammers" and Sui and Move as offering "laser cutters and welders."

The governance token is where the reporting gap is widest. The mint is SLNDpmoWTVADgEdndyvWzroNL7zSi1dF9PC3xHGtPwp, and three sources give three different supplies:

• The chain: getTokenSupply on mainnet returns 3,334,462.284268 SLND. Jupiter's token API independently reports the same figure as both total and circulating supply, with a market cap of $659,019 and 11,687 holders.
• Listing aggregators: around 45.66 million, self-reported by the project, following the May 2025 burn of 50 million tokens — half the supply — which was covered at the time by ChainCatcher and others.
• CoinGecko: total supply 100,000,000, circulating supply 0, market cap $0. Its price feed works ($0.1977) but its supply data is the pre-burn figure from the documentation.

We are reporting the on-chain figure because it is the only one two independent sources agree on, and because the documented 100M number has a clear provenance: the Token page still reads "There are 100M SLND tokens," distributed 60% community, 25% core team, 15% VCs and angels, and it is marked "Last updated 3 years ago." We cannot date the gap between 45.66 million and 3.33 million from public sources, and we are not going to guess at it.

Two further facts about the mint are worth stating plainly. The mint authority is not null. It is set to EaFPY9LTQeFR7SEyfbKKFuVMtYvUBbYiiK7WvBJJ7iBU, which the DAO page labels "treasury msig," so new SLND can still be issued. The freeze authority is null, which is the good outcome — nobody can freeze your tokens.

And the liquidity is almost gone. DexScreener shows three pairs holding $8,177 between them — Orca SLND/USDC at $4,301, Raydium SLND/USDC at $3,616, Meteora SLND/bSOL at $260 — on $42.99 of 24-hour volume across 28 trades. Against the on-chain market cap that is 1.24% of the float sitting in pools, and the whole token is worth 0.78% of the deposits it governs. CoinGecko puts the token 98.81% below its $16.65 all-time high of 7 November 2021.

There is a data-quality tell inside DexScreener's own response, too. The Orca and Raydium pairs report a market cap of about $17.9 million, which back-solves to roughly 90.5 million tokens; the Meteora pair reports $662,623, which back-solves to the real 3.33 million. Same token, same minute, two answers — which is why the rule is to take supply from getTokenSupply and never from a listing site. The same discipline applies to any pool you are about to provide liquidity into, where the figure that actually determines your outcome is impermanent loss rather than the advertised APY.

Three practical conclusions, none of which require you to have an opinion about whether Save is a good protocol.

The documentation is not the product. Everything above was found by checking what the chain and the live app say against what the docs say. On Solana that check is cheap: a program ID gives you the owner, the program data account, the upgrade authority and the slot of the last upgrade in two RPC calls, and verify.osec.io tells you in one more whether the deployed binary matches the public source. Run it before you deposit anywhere. We ran the same sequence on Sanctum and on Solana's MEV infrastructure, and it has never once failed to turn up something the marketing page omits.

A rebrand leaves a residue, and the residue is a security problem. Two live front ends, a documentation set at a dead subdomain ranking first on Google, a bug bounty page that 404s on the new domain and 400s on the old one, and a security page whose only dated audit row reads September 28, 2021. Each of those is small. Together they are the reason a user ends up on the wrong site.

Custody and verifiability are not the same question. Save is non-custodial in the sense that matters — your collateral is yours, and the docs are explicit that the lending market owner "can only change the configs and can't move funds from the user." It is also upgradeable by a single key with no timelock, which is a different kind of exposure, and one that the 2022 whale vote showed is not theoretical. Both statements are true at once, and a protocol that tells you so in its own documentation is being more honest than most.

That distinction is the one we care about most at uwuu. Copy trading has exactly the same shape of problem: you can be non-custodial and still be asked to trust a leaderboard you cannot audit. Our answer is a verified on-chain leaderboard — every trader's performance is reconstructed from their public wallet history, so the number you are copying is one you can check yourself with the same RPC calls used throughout this article, and uwuu's fee is performance-based, so you pay only when a copied trade makes you money.

Is Solend still active?

Yes. Solend rebranded to Save in July 2024 and the lending program is still live on Solana mainnet, holding $84.5 million of net TVL against $159.1 million borrowed as of 17 October 2026. The current app is at save.finance; the old solend.fi app also still returns HTTP 200.

What is Solend's program ID?

The main lending program is So1endDq2YkqhipRh3WViPa8hdiSpxWy6z3Z6tMCpAo. Its program data account is DMCvGv1fS5rMcAvEDPDDBawPqbDRSzJh2Bo6qXCmgJkR, and the live upgrade authority on 17 October 2026 is RY93CZYe5g6drtG7W9PmHRPzaBLZ1uwihTzayQTmJfh — which is not either of the two addresses the documentation gives.

Does Solend have flash loans?

There is a limited flash loan instruction in the deployed program, but the composable version the docs describe was never shipped. The Flash Loans page, last updated four years ago, links to pull request #70 for the improved implementation; that pull request was closed in January 2023 without being merged.

What is Solend's flash loan fee?

The flash loan fee is a per-reserve parameter rather than a single protocol-wide rate, so it has to be read in the app for the specific pool. For ordinary borrowing, the published figures are a 20% interest rate spread, an origination fee of 10 basis points on most vaults split 80/20 between protocol and host, and a liquidation fee of 30% of the liquidation penalty.

What happened to the Solend whale in 2022?

Nothing, in the end. Proposal SLND1 granted Solend Labs emergency power to take over the account via a smart contract upgrade; SLND2 invalidated it a day later. SOL recovered, the position was never liquidated and the wallet was never seized. Its last on-chain transaction was 18 March 2024.

Is SLND the same token as SAVE?

The on-chain mint is still SLNDpmoWTVADgEdndyvWzroNL7zSi1dF9PC3xHGtPwp and Jupiter still lists it as SLND. Its on-chain supply is 3,334,462 tokens with 11,687 holders and about $8,177 of total DEX liquidity, which is a fraction of the figures listing sites carry — take supply from the chain, not from an aggregator.