Spark scores 6 points above Compound, led by oracle liquidation price, blast radius, and rate authorship. Spark ranks #2 and Compound #6 of 8 published lending & borrowing. Every term of the gap is on the ledger below; the inputs are on the audit receipt.
Spark80
vs
Compound74
Scores are out of 100; published scores fall within 41–99 · ranks #2 and #6 of 8
Spark scores 6 points above Compound, led by oracle liquidation price, blast radius, and rate authorship.
Spark #2 of 880
Compound #6 of 874
Scores are out of 100; published scores fall within 41–99; a bar is empty at 41 and full at 99.
Why, term by termSpark leads on 3 of 4 published criteria; Compound on 1. The terms sum to the gap.
TermSparkCompoundWeightPoints
Oracle liquidation price8/106/107.7+4.0for Spark+2.0 of 32
Blast radius8/107/105.8+1.5for Spark+0.8 of 32
Rate authorship10/109/105.8+1.5for Spark+0.8 of 32
Shortfall record7/108/104.8−1.2for Compound−0.6 of 32
Net of the terms+5.7
Reproduced gap+5.7
Published gapAfter score rounding and bounds — see the note below8074+6.0
Weight is in points of the 86-point weighted total. Points are score points of the gap between the two.
SparkScore 80SparkLend is an Ethereum-only Aave v3 fork run as Sky's in-house money market: prices come from a three-oracle median of RedStone, Chainlink and Chronicle with a Uniswap TWAP as last resort, your deposit is never re-routed by a curator, and the borrow rate is an immutable per-reserve IRM — SSR-anchored on the stablecoin reserves — that Sky governance can only change by redeploying the contract.Strength: Rate authorship — 10/10Watch-out: Read the oracle-type field on your collateral in the app before you post it, and read the liquidity line before you supply. SparkLend does not price every asset at market: 'stablecoins such as USDS and DAI use a fixed price of 1.00 USD, which does not deviate even if the asset's market price moves away from this value' (https://docs.spark.fi/products/sparklend/guides/oracles). If you post USDS or DAI as collateral, a real-world depeg will not move your health factor at all — and the peg-ratio killswitch that exists for exactly this failure covers only 'wstETH, rETH, weETH, cbBTC, WBTC, LBTC', not the fixed-price stablecoins (https://docs.spark.fi/products/sparklend). USDS is issued by Sky, whose governance both sets that $1.00 constant and authors your borrow rate, so the protection and the exposure share a controller. Second, the USDC supply rate you are being shown was deliberately engineered by thinning the withdrawal buffer, and Spark's own risk body said so before the vote: 'At 98% utilization, only about 2% of supplied USDC is immediately available before new deposits, repayments, liquidations, or other liquidity sources are needed. Withdrawals become more fragile', and 'The proposal improves the displayed supply rate by accepting a thinner liquidity buffer' (Spark Risk Council, 2026-06-22, https://forum.skyeco.com/t/july-2-2026-proposed-changes-to-spark-for-upcoming-spell/27982).CompoundScore 74COMP-governed onchain money market whose Compound III (Comet) design gives each market exactly one borrowable base asset and a set of capped, non-lent collateral assets, priced by a single named Chainlink feed per asset with no published deviation bound, and liquidated whole by an absorb call paid out of protocol reserves.Strength: Rate authorship — 9/10Watch-out: Compound now has one of the cleanest RATE disclosures in this pool and one of the thinnest ORACLE disclosures, and a buyer who conflates the two will misprice it. Four specific traps. First, the page that tells you what liquidates you — https://docs.compound.finance/liquidation/ — never names an oracle at all ("This function uses the price returned by the protocol’s price feed"); Chainlink is named two clicks away on the helper-functions page, and a machine check of all nine Compound doc pages returns zero occurrences of deviation, heartbeat, staleness, median, fallback, killswitch or price bound. The one bounded feed Compound ever ran — the multisig-settable min/max and constant-price override added to rsETH after the 18 April 2026 Kelp bridge exploit — was REVERTED in May 2026 via Proposal 586, so as of today no Compound market carries a numeric price bound. Second, the docs registry is a snapshot, not live state: stamped 2026-02-28, it still advertises rsETH on the Ethereum USDC Comet at 85% borrow CF and a 1,500-token supply cap when the on-chain cap has since been set to 0, and it still lists ten mainnet networks when six of them — Polygon, Unichain, Linea, Mantle, Ronin and Scroll — are approved for deprecation with supply caps zeroed, collateral factors halved and supply rates being set to 0%. Read the Comet and the forum, not the docs table. Third, the one-week governance timelock that makes Compound's rate authorship trustworthy does NOT cover the emergency path: the Community Multi-Sig pause guardian can freeze withdrawals instantly (it did so across six networks in April 2026), and a 0% collateral factor mechanically locks collateral until debt is fully repaid — both disclosed on the forum, neither in the docs. Fourth, the named backstop is real but shrinking and undocumented at the tail: reserves on the Ethereum WETH Comet fell from 921.22 to 908.46 WETH between 10 May and 5 August 2026 because the comet "retains none of the interest borrowers pay at its optimal utilization", no Compound page says what happens once reserves are exhausted, and Compound discloses its losses well on comp.xyz and not at all on either page titled Security.
Quick facts
At a glance
3 of 4 rows read the same for both.
FactSparkCompound
PricingSnapshotSparkLend's stablecoin reserves price off the Sky Savings Rate (SSR) plus a fixed spread up to the kink, and suppliers receive the borrow interest less a 10% reserve factor; there is no deposit or curator fee. The 10 September 2026 spell proposal (https://forum.skyeco.com/t/september-10-2026-proposed-changes-to-spark-for-upcoming-spell/28208, posted 31 August 2026, fetched 28 September 2026) reads the rate on-chain at Ethereum block 25,861,675 (29 August 2026): "getVariableRateSlope1() returns 3.9594644783474038423088%, which equals the SSR of 3.4594644783474038423088% plus the 0.50% spread exactly." For USDT, the largest reserve, it cuts that spread "from 0.005e27 (+0.50%) to 0.001e27 (+0.10%)" and states the effect: "The borrow rate at the kink falls from 3.9595% to 3.5595%" and "At 95% utilisation and a 10% reserve factor that is a supply rate of about 3.04%"; at the time of posting "The live borrow rate is 3.4824% and the live supply rate 2.6186%." The same post gives the other stablecoin curves: "USDC runs a 98% kink with a +1.00% spread". These are SSR-linked and move with every SSR change and with utilisation; they are not a fixed price. The 2 July 2026 SSR figure this note previously carried is superseded.Compound III charges no subscription, listing or curator fee; the protocol's take is the spread between separate supply and borrow curves that governance sets. The Interest Rates page (https://docs.compound.finance/interest-rates/, fetched 28 September 2026): "Users with a positive balance of the base asset earn interest, denominated in the base asset, based on a supply rate model; users with a negative balance pay interest based on a borrow rate model. These are separate interest rate models, and set by governance." Compound publishes the size of that spread only per proposal on its forum. For the Ethereum WETH Comet, Gauntlet's 5 August 2026 post (https://www.comp.xyz/t/ethereum-weth-comet-borrow-rate-irm-updates/7999, fetched 28 September 2026) states "Compound v3 has no reserve factor parameter" and prints the implied reserve factor as "0.00%" at the kink and "10.00%" at full utilisation, adding "The comet currently retains none of the interest borrowers pay at its optimal utilization." Comets being wound down are repriced by the same route: a 2 September 2026 forum proposal sets the Ethereum wstETH Comet's supply rate to zero ahead of deprecation. Live per-market APYs are on-chain only, so no APY is quoted.
Asset coverageEditorial——
Region fitSnapshotglobalglobal
Custody modelEditorialNon-custodial: no company holds the coinsNon-custodial: no company holds the coins
From the published criteria
Which should you pick?
Spark
80#2 of 8
Leads
Pick it when
Pick Spark if oracle liquidation price matters most — it scores 8/10 to Compound's 6/10 there (weight 32).
Strength
Rate authorship — 10/10
Tradeoff
Read the oracle-type field on your collateral in the app before you post it, and read the liquidity line before you supply. SparkLend does not price every asset at market: 'stablecoins such as USDS and DAI use a fixed price of 1.00 USD, which does not deviate even if the asset's market price moves away from this value' (https://docs.spark.fi/products/sparklend/guides/oracles). If you post USDS or DAI as collateral, a real-world depeg will not move your health factor at all — and the peg-ratio killswitch that exists for exactly this failure covers only 'wstETH, rETH, weETH, cbBTC, WBTC, LBTC', not the fixed-price stablecoins (https://docs.spark.fi/products/sparklend). USDS is issued by Sky, whose governance both sets that $1.00 constant and authors your borrow rate, so the protection and the exposure share a controller. Second, the USDC supply rate you are being shown was deliberately engineered by thinning the withdrawal buffer, and Spark's own risk body said so before the vote: 'At 98% utilization, only about 2% of supplied USDC is immediately available before new deposits, repayments, liquidations, or other liquidity sources are needed. Withdrawals become more fragile', and 'The proposal improves the displayed supply rate by accepting a thinner liquidity buffer' (Spark Risk Council, 2026-06-22, https://forum.skyeco.com/t/july-2-2026-proposed-changes-to-spark-for-upcoming-spell/27982).
Compound
74#6 of 8
Pick it when
Pick Compound if shortfall record matters most — it scores 8/10 to Spark's 7/10 there (weight 20).
Strength
Rate authorship — 9/10
Tradeoff
Compound now has one of the cleanest RATE disclosures in this pool and one of the thinnest ORACLE disclosures, and a buyer who conflates the two will misprice it. Four specific traps. First, the page that tells you what liquidates you — https://docs.compound.finance/liquidation/ — never names an oracle at all ("This function uses the price returned by the protocol’s price feed"); Chainlink is named two clicks away on the helper-functions page, and a machine check of all nine Compound doc pages returns zero occurrences of deviation, heartbeat, staleness, median, fallback, killswitch or price bound. The one bounded feed Compound ever ran — the multisig-settable min/max and constant-price override added to rsETH after the 18 April 2026 Kelp bridge exploit — was REVERTED in May 2026 via Proposal 586, so as of today no Compound market carries a numeric price bound. Second, the docs registry is a snapshot, not live state: stamped 2026-02-28, it still advertises rsETH on the Ethereum USDC Comet at 85% borrow CF and a 1,500-token supply cap when the on-chain cap has since been set to 0, and it still lists ten mainnet networks when six of them — Polygon, Unichain, Linea, Mantle, Ronin and Scroll — are approved for deprecation with supply caps zeroed, collateral factors halved and supply rates being set to 0%. Read the Comet and the forum, not the docs table. Third, the one-week governance timelock that makes Compound's rate authorship trustworthy does NOT cover the emergency path: the Community Multi-Sig pause guardian can freeze withdrawals instantly (it did so across six networks in April 2026), and a 0% collateral factor mechanically locks collateral until debt is fully repaid — both disclosed on the forum, neither in the docs. Fourth, the named backstop is real but shrinking and undocumented at the tail: reserves on the Ethereum WETH Comet fell from 921.22 to 908.46 WETH between 10 May and 5 August 2026 because the comet "retains none of the interest borrowers pay at its optimal utilization", no Compound page says what happens once reserves are exhausted, and Compound discloses its losses well on comp.xyz and not at all on either page titled Security.
Self-identify in 10 seconds
Which one is actually for you?
SparkLeads
Choose Spark if oracle liquidation price matters most
Oracle liquidation price: 8/10 against 6/10 for Compound (weight 32)
Rate authorship: 10/10 against 9/10 for Compound (weight 24)
Blast radius: 8/10 against 7/10 for Compound (weight 24)
SparkScore 80SparkLend is an Ethereum-only Aave v3 fork run as Sky's in-house money market: prices come from a three-oracle median of RedStone, Chainlink and Chronicle with a Uniswap TWAP as last resort, your deposit is never re-routed by a curator, and the borrow rate is an immutable per-reserve IRM — SSR-anchored on the stablecoin reserves — that Sky governance can only change by redeploying the contract.Strength: Rate authorship — 10/10Watch-out: Read the oracle-type field on your collateral in the app before you post it, and read the liquidity line before you supply. SparkLend does not price every asset at market: 'stablecoins such as USDS and DAI use a fixed price of 1.00 USD, which does not deviate even if the asset's market price moves away from this value' (https://docs.spark.fi/products/sparklend/guides/oracles). If you post USDS or DAI as collateral, a real-world depeg will not move your health factor at all — and the peg-ratio killswitch that exists for exactly this failure covers only 'wstETH, rETH, weETH, cbBTC, WBTC, LBTC', not the fixed-price stablecoins (https://docs.spark.fi/products/sparklend). USDS is issued by Sky, whose governance both sets that $1.00 constant and authors your borrow rate, so the protection and the exposure share a controller. Second, the USDC supply rate you are being shown was deliberately engineered by thinning the withdrawal buffer, and Spark's own risk body said so before the vote: 'At 98% utilization, only about 2% of supplied USDC is immediately available before new deposits, repayments, liquidations, or other liquidity sources are needed. Withdrawals become more fragile', and 'The proposal improves the displayed supply rate by accepting a thinner liquidity buffer' (Spark Risk Council, 2026-06-22, https://forum.skyeco.com/t/july-2-2026-proposed-changes-to-spark-for-upcoming-spell/27982).
Compound
Choose Compound if shortfall record matters most
Shortfall record: 8/10 against 7/10 for Spark (weight 20)
CompoundScore 74COMP-governed onchain money market whose Compound III (Comet) design gives each market exactly one borrowable base asset and a set of capped, non-lent collateral assets, priced by a single named Chainlink feed per asset with no published deviation bound, and liquidated whole by an absorb call paid out of protocol reserves.Strength: Rate authorship — 9/10Watch-out: Compound now has one of the cleanest RATE disclosures in this pool and one of the thinnest ORACLE disclosures, and a buyer who conflates the two will misprice it. Four specific traps. First, the page that tells you what liquidates you — https://docs.compound.finance/liquidation/ — never names an oracle at all ("This function uses the price returned by the protocol’s price feed"); Chainlink is named two clicks away on the helper-functions page, and a machine check of all nine Compound doc pages returns zero occurrences of deviation, heartbeat, staleness, median, fallback, killswitch or price bound. The one bounded feed Compound ever ran — the multisig-settable min/max and constant-price override added to rsETH after the 18 April 2026 Kelp bridge exploit — was REVERTED in May 2026 via Proposal 586, so as of today no Compound market carries a numeric price bound. Second, the docs registry is a snapshot, not live state: stamped 2026-02-28, it still advertises rsETH on the Ethereum USDC Comet at 85% borrow CF and a 1,500-token supply cap when the on-chain cap has since been set to 0, and it still lists ten mainnet networks when six of them — Polygon, Unichain, Linea, Mantle, Ronin and Scroll — are approved for deprecation with supply caps zeroed, collateral factors halved and supply rates being set to 0%. Read the Comet and the forum, not the docs table. Third, the one-week governance timelock that makes Compound's rate authorship trustworthy does NOT cover the emergency path: the Community Multi-Sig pause guardian can freeze withdrawals instantly (it did so across six networks in April 2026), and a 0% collateral factor mechanically locks collateral until debt is fully repaid — both disclosed on the forum, neither in the docs. Fourth, the named backstop is real but shrinking and undocumented at the tail: reserves on the Ethereum WETH Comet fell from 921.22 to 908.46 WETH between 10 May and 5 August 2026 because the comet "retains none of the interest borrowers pay at its optimal utilization", no Compound page says what happens once reserves are exhausted, and Compound discloses its losses well on comp.xyz and not at all on either page titled Security.
Criterion by criterion
Scorecard
Published criteria · weight in points of 868 of 8 cells scored. Strength is read from the evidence file; the weight is the published one. Methodology
Oracle liquidation price
by 2 of 10weight 7.7 of 86Spark leads
Spark8/10 out of 10
Compound6/10 out of 10
Rate authorship
by 1 of 10weight 5.8 of 86Spark leads
Spark10/10 out of 10
Compound9/10 out of 10
Blast radius
by 1 of 10weight 5.8 of 86Spark leads
Spark8/10 out of 10
Compound7/10 out of 10
Shortfall record
by 1 of 10weight 4.8 of 86Compound leads
Spark7/10 out of 10
Compound8/10 out of 10
Score dimensions
weights sum to 86
The published score is five dimensions, each weighted for the category. "Published criteria" is the criteria block above; the other four are the engine's own terms — availability, friction, trust record, measured data. Each bar is one side's points out of that dimension's maximum; the weight beside the name is its points of the 86-point weighted total. The published criteria above can fill 24 of the block’s 30. A weight marked “constant here” belongs to a dimension every ranked product in the category holds at the same value: it still counts in the total, and it orders nothing here.
Availabilityweight 20 of 86 · constant here
Spark14/20Compound14/20
Tie
Published criteriaweight 30 of 86
Spark19.9/30Compound17.7/30
Spark leads
Frictionweight 10 of 86 · constant here
Spark2.7/10Compound2.7/10
Tie
Trust recordweight 8 of 86 · constant here
Spark3.2/8Compound3.2/8
Tie
Measured dataweight 18 of 86 · constant here
Spark8.3/18Compound8.3/18
Tie
Decision rulePublished order
Choose Spark unless the term Compound leads on is the one you are buying.
Pick Compound if shortfall record matters most — it scores 8/10 to Spark's 7/10 there (weight 20).
The published order weighs every criterion the same way for everyone. To weigh them your way, run .
From the published order
Next best action
If this isn't your case
Other matchups worth comparing
How you can lose money here, and what the ranking scores
Control failure: a key or vote changes the rules, freezes or mints.
Scored by Who sets your rate — a utilisation curve or a curator? · 24% (Spark 10/10, Compound 9/10)
Oracle failure: a wrong price or outcome liquidates or settles a position.
Scored by What price gets you liquidated, and how far can it drift? · 32% (Spark 8/10, Compound 6/10)
Insolvency failure: the firm or issuer that owes you cannot pay.
Scored by If one market goes bad, how much of yours goes with it? · 24% (Spark 8/10, Compound 7/10)
Scored by Has it lost lender money before, and did it publish it? · 20% (Spark 7/10, Compound 8/10)
Not assessed by this ranking: Contract failure, Regulatory failure, Execution failure, Exit failure. A high place says nothing about these; the Methodology tab says what each means.
Questions
Frequently asked questions
Spark scores 6 points above Compound, led by oracle liquidation price, blast radius, and rate authorship. Published scores, out of 100: Spark 80 (rank 2 of 8 lending & borrowing), Compound 74 (rank 6 of 8).
Reviewed means both products compared here was read against its sources on or after that date; every ranked category is re-read at least every 45 days. Models research, extract and parse. A deterministic engine ranks. Every quote passes an adversarial refute pass.
Answer a few quick questions and get one clear recommendation based on how you actually plan to use crypto — with rationale, tradeoffs, and a path to monitor the decision.
No signup·Free first pass · Private·No spam · No account
Not financial advice · Independent · Always do your own research
Not financial advice · For informational purposes only · Always do your own research
