Ratings · Measured, not sponsored
Lending platforms
Every comparison ranks these protocols by how much money is parked in them. We asked two better questions instead: when this market last broke, did the depositors get paid — and is anybody actually borrowing the capital you are being asked to supply?
$18.39bn deposited and no depositor-loss event in its history. You pay for that in yield: 2.03% gross.
Lost $197m in 2023 and returned 100% of it. Also the highest-earning market we measured, at 6.37% gross.
Nine figures of bad debt from a six-hour oracle manipulation in 2021, never recovered — and the second-lowest yield on the page.
What we did · August 27, 2026We took every incident in the category’s history and asked what happened to the depositors afterwards, not just what the loss headline said. Then we computed what each market genuinely earns per dollar deposited, which is a number no front end shows you.
“IT GOT HACKED” IS NOT A FACT YOU CAN RANK ON
One protocol lost nine figures and made everyone whole. One was never breached and cost its depositors nine figures. The incident record only means something once you attach what happened to the people who had money in it — which is why that question carries 32% of the score here.
Is anybody actually borrowing your money?
Thirty days of protocol fees, annualised, per dollar deposited. Not an advertised APY — the gross interest the market really generated.
JustLend holds the fourth-largest deposit base in the category and generates one eleventh of Euler’s rate on it. That is what parked collateral looks like from the inside, and no TVL league table will ever show it to you.
The ranking
Click any row for the quick read, or open a company for its full profile and per-criterion scores. Measured August 27, 2026. How we rate.
HOW WE SCORED THIS — CRITERIA AND WEIGHTS+
A lending market is a promise that your deposit can be withdrawn. Every failure in this category’s history is that promise breaking in one of two ways: a contract drained, or a collateral price manipulated until the loans it backed became bad debt. So the score is led by what has already gone wrong at each protocol and — the question almost nobody asks — whether the depositors got their money back.
Every exploit and every bad-debt event with dates and amounts, and then the question that decides whether it matters: were depositors made whole, and by whom. Euler lost $197m in March 2023 and returned 100% of it after negotiating with the attacker. Venus carried nine figures of bad debt from a single oracle manipulation in 2021. Those are opposite outcomes from superficially similar headlines, and a rating that treats "has been hacked" as one fact is useless.
Computed by us: thirty days of protocol fees, annualised, divided by total value locked. That is the gross interest the market genuinely generated per dollar deposited — before the protocol’s cut and before incentives. It is the honest version of the APY a front end shows you, and it separates a market where borrowers are really paying from one where the TVL headline is parked capital earning close to nothing.
Total value locked at a stated timestamp, and whether the market is deep enough that a withdrawal at size does not itself move the utilisation rate. In lending, depth is not a convenience — a market that is 99% utilised cannot pay you out until somebody repays.
Sourced to documentation: are markets isolated or pooled, what oracle secures each collateral, who can change a listing or a loan-to-value ratio and how fast, and — for the newer curated-vault designs — who exactly is picking your risk and whether they carry any of it. A single shared pool means one bad listing can impair everyone.
Deployments with real deposits, not deployment count. A protocol on forty chains with liquidity on three is scored on the three.
Weights sum to 100. If we cannot verify a criterion, we delete it rather than score it on impressions — read the methodology.
Is there a free alternative?
If you are lending stablecoins for yield, compare it against a Treasury bill and be honest about the answer.
The gross rates our measurement found run from 0.56% to 6.37% a year, and the ones at the top come with smart-contract risk, oracle risk, curator risk and the possibility of a market at 99% utilisation that cannot pay you out this week. A government bill pays a competitive rate with none of those. That is not an argument against using these protocols — it is an argument for knowing what the spread is actually compensating you for, and for treating anything above the risk-free rate as the price of the risks listed on this page rather than as free money.
What changed since last time
- 2026-08-27Category published.Every comparison lists the same protocols by TVL. TVL says how much money is parked, not whether it earns anything or whether the last time this protocol broke the depositors got paid.
Questions
Which crypto lending platform is safest?+
Aave, on the evidence: $18.39bn deposited, the deepest market in the category, and no depositor-loss event in its history. That safety is bought with slow, contested listings and conservative parameters, and it costs yield — our measurement puts its gross rate at 2.03% a year per dollar deposited, well under the smaller markets. Safety in this category is not a feature you get for free; it is a rate you accept.
Does “this protocol was hacked” tell me it is unsafe?+
On its own, no — and this category contains the proof. Euler lost $197m in March 2023 and got 100% of it back over three weeks by negotiating with the attacker, who eventually apologised on-chain and returned roughly $240m of assets. Venus was never hacked at all: someone moved the price of its own governance token from $76 to $143 and back within six hours, borrowed against it at the peak, and left more than $100m of bad debt that was never recovered. Same headline shape, opposite outcome for depositors. The question is never whether something went wrong — it is who paid for it.
What does the gross yield figure on this page mean?+
It is ours, computed rather than quoted: thirty days of protocol fees, annualised, divided by total value locked. That is the gross interest each market genuinely generated per dollar deposited, before the protocol’s cut and before token incentives. It answers a question front-end APYs do not — whether anyone is actually borrowing. The spread is enormous: 6.37% at Euler, 0.56% at JustLend, on markets that both look large.
Are curated vaults safer than a shared lending pool?+
They move the risk rather than remove it. Isolated markets genuinely stop one bad collateral from contaminating everything else, which is a real improvement over the shared-pool designs where most of this category’s bad debt came from. But most depositors do not use the primitive directly — they deposit into a vault where a third-party curator chooses which isolated markets their money enters and at what loan-to-value ratio, usually with no capital of their own at stake. If you use one, read the allocations.
Is lending stablecoins in DeFi worth the risk?+
Compare it honestly with a Treasury bill before you decide. The gross rates we measured run from 0.56% to 6.37%, and the higher ones come with smart-contract risk, oracle-manipulation risk, curator risk and the possibility of a market too utilised to pay you out this week. Treat anything above the risk-free rate as the price of those specific risks, not as free money — and size the position as if the worst case on this page could happen to you.
How this is funded
It is not. There are no affiliate links on this page or anywhere on this site, no paid placements, and no sponsored positions. Nobody in this table can buy a place in it, accelerate their inclusion, or influence a score — and none of them paid us anything, because there is nothing here to buy. The full policy.