Chainlink Feeds vs Pyth Feeds: Delivery

Pull-based, on-chain verification: Data is stored on-chain and fetched by contracts via latestRoundData. Every update is cryptographically signed by a decentralized network (e.g., 31+ nodes for ETH/USD). This matters for protocols requiring maximum security and censorship resistance, like lending platforms (Aave, Compound) or reserve-backed stablecoins.

Higher latency, predictable cost: Updates are pushed on-chain at regular intervals (e.g., every 1-24 hours). While this can mean slower price discovery, gas costs are borne by the oracle network, not the dApp. This matters for non-latency-sensitive, cost-predictable applications like insurance protocols or treasury management.

Push-based, cross-chain streaming: Price updates are published to a permissioned Pythnet, then relayed as signed messages to consumer chains via Wormhole. Contracts receive data directly into storage. This matters for high-frequency trading, perpetuals, and options where sub-second latency is critical (e.g., Hyperliquid, Drift Protocol).

User-paid updates, maximum freshness: dApps pay to pull the latest signed price from the Pyth contract, ensuring they get the most recent data. This creates a pay-for-freshness model ideal for derivatives and spot DEXes where being first to an accurate price is worth the gas (e.g., Synthetix, MarginFi).

On-demand data retrieval: Users pay gas only when they request an update, not for every new price. This matters for low-frequency protocols like insurance, options, or governance, where data updates may be needed only weekly or monthly, optimizing operational costs.

Deterministic data age: The updatedAt timestamp is on-chain, allowing contracts to verify exactly how stale the data is and enforce freshness thresholds. This matters for risk-sensitive DeFi like lending (Aave, Compound) where using outdated prices can lead to undercollateralized positions.

Continuous push updates: Data is written on-chain with every price change (e.g., on Solana, Sui) or via Wormhole to other chains, achieving sub-second latency. This matters for high-frequency trading (DEXs, perpetuals) and liquid staking derivatives where arbitrage depends on millisecond-fresh prices.

Requires active management: If a contract doesn't call latestRoundData() frequently, it may read outdated data. This matters for automated systems that must budget for and schedule update transactions, adding operational overhead and potential points of failure.

Costs are socialized: While users don't pay per update, the network pays for continuous writes. This can lead to higher base protocol costs or inflation, which matters for cost-sensitive applications on high-throughput chains where write costs are a significant portion of fees.

Direct on-chain updates: Publishers push price updates directly to the Pythnet appchain, which are then relayed to consumer chains via Wormhole. This enables sub-second price updates critical for high-frequency DeFi (e.g., perpetuals on Hyperliquid, MarginFi). This matters for protocols where stale data directly equates to arbitrage losses.

Cost borne by publishers: Data providers pay the gas to publish updates to Pythnet, insulating consumer protocols from volatile on-chain gas fees. This creates a predictable, near-zero operational cost for dApps reading prices. This matters for scaling high-throughput applications on Solana or Avalanche where gas spikes can cripple pull-model updates.

Consumer-initiated updates: DApps or keepers explicitly request an update, guaranteeing the data is fresh at the exact moment of their transaction (e.g., a liquidation). This eliminates reliance on a push schedule. This matters for low-liquidity or volatile assets where you cannot trust a recently-pushed value, and for secure settlement layers like Ethereum L1.

Decoupled data retrieval: The oracle network updates an on-chain Aggregator contract only when paid to do so. This simplifies security audits (state changes are explicit) and aligns with Ethereum's pull-centric design. This matters for protocols prioritizing battle-tested, minimal trust assumptions over absolute speed, such as money markets (Aave) or reserve-backed stablecoins.