| Author | Pyth Data Association |
|---|---|
| Version / date | Version 2.0 / September 28, 2023 |
| Project | Pyth Network |
| License | No license is stated in the linked document |
| Primary source | Official Pyth Network Whitepaper v2.0 (PDF) |
This page summarizes and analyzes a technical document for educational purposes; it is not investment or financial advice. The whitepaper reviewed here was published in 2023. Its descriptions of Pythnet and Wormhole explain the design at that time and should not be read as a complete description of Pyth’s infrastructure after the 2026 migration.
Pyth: A First-Party Financial Oracle Network, version 2.0, was published by the Pyth Data Association on September 28, 2023. Its central idea is to collect prices directly from firms and market participants with access to pricing sources, then make that data available to on-chain and off-chain applications. Version 2.0 updated Pyth’s earlier Solana-focused design with a cross-chain delivery model built around Pythnet and Wormhole.
Publication and Context
An oracle carries information from outside a blockchain into smart contracts. In financial protocols, stale or unrepresentative prices can affect collateral values, liquidations, and settlement. Pyth emphasizes first-party publishers: exchanges, market makers, and financial firms publish prices they collect themselves. That differs from a simple picture of an aggregator that only fetches public APIs, but it does not mean every other oracle product uses the same method.
The v2.0 whitepaper distinguishes publishers from consumers. Publishers contribute prices and confidence intervals; in the design described in the document, publishers can receive fees while consumers pay to update prices. This is the 2023 economic model in the paper, not proof that each of those fees and incentives still applies in 2026.
Pythnet and Pull Delivery in the Whitepaper
The 2023 design uses Pythnet, a Solana-based application blockchain, to store publisher updates and calculate aggregate prices. Pythnet validators create Merkle roots for updated prices and distribute those roots across chains through a messaging network such as Wormhole. On a destination chain, a consumer requests the updates it needs, and the receiving contract checks the proof and signatures before storing the price. Because the application requests an update when needed, this is called a pull model.
Pull delivery means transactions on the destination chain are submitted when an application actually sends an update, but it also makes freshness checks the integrator’s responsibility. If an application reads a value without requesting a recent update, that value can lag the market. The consuming contract or application needs clear rules for data age and what to do when a feed is not fresh enough.
Price Aggregation and Confidence
In the version 2.0 algorithm, each publisher has a stake-based weight. For the price calculation, each publisher contributes three votes: its reported price, that price plus its confidence interval, and that price minus its confidence interval. The algorithm takes a stake-weighted median; the aggregate confidence then summarizes how publisher prices are distributed around the aggregate. The whitepaper does not say publisher weights are inversely proportional to confidence-interval width, as this page previously claimed.
Current Pyth Pro documentation describes a newer aggregation process: the aggregate price is the median of publisher prices, while confidence is calculated from the distribution of publisher prices, bids, and asks. Confidence measures disagreement among publisher observations; it is not automatically the same as market volatility. Current documentation also describes an EMA with a window of about one hour, not a 10-second half-life. Because the design and products have changed, figures and algorithms from the 2023 document need dates and should not be presented as current specifications.
Infrastructure Changes Since 2023
In 2026, the DAO approved OP-PIP-100 to migrate Pyth Core users to Pyth Pro infrastructure and retire Pythnet. Pyth’s documentation says the Core upgrade was completed on August 26, 2026. For upgraded Core compatibility, five independently operated routers sign Merkle roots, with a documented threshold of three out of five signatures. This replaces the older arrangement in which Pythnet and 13 of 19 Wormhole Guardians were signers. Hermes now requires an API key. Support and migration details can vary by chain and feed.
As of October 10, 2026, a Pyth developer-forum notice scheduled the deprecation of certain EVM Push Feeds for October 15, 2026. The notice applies to the feeds and chains it lists; it does not mean every Pyth feed is being retired. Check the latest notice before relying on a specific feed.
Pyth Core and Pyth Pro Today
Pyth Core is the on-chain oracle interface built for DeFi integrations. Pyth Pro is a commercial data service with authenticated access, feed options, and configurable delivery channels and schedules. The DAO approved moving Core’s infrastructure toward Pyth Pro while retaining a compatibility layer for many Core interfaces. Treating Core and Pro as two entirely separate networks can therefore be misleading: product names, access, fees, and support differ while the infrastructure is being consolidated.
The v2.0 whitepaper remains useful for understanding why Pyth initially emphasized first-party data and pull delivery. For today’s integration requirements, consult current Pyth Core and Pyth Pro documentation alongside the 2023 PDF.
Fees, Incentives, and Limitations
The whitepaper’s fee model needs to be separated from current conditions. Official documentation says Pyth Core update fees are now zero on supported EVM mainnets, following OP-PIP-128 as part of the Core sunset. Pyth Pro, as a subscription service, has its own pricing. Oracle Integrity Staking (OIS) is separate from product fees: its documentation says the OIS reward rate was set to zero in April 2026, while staking and slashing remain active. Broad claims that publishers are always paid per update or that all Pyth access is free are not supported by these sources.
First-party data can reduce reliance on one type of intermediary, but the resulting feed still depends on its sources, publishers, aggregation, and integration. A signed price is not a guarantee that it is always fresh or suitable for every decision. Integrators should check timestamps, publisher counts, confidence values, and feed behavior when updates are delayed.
Reality Check
The 2023 whitepaper documents a design and economic assumptions from that period. By 2026, the transition away from Pythnet, the end of Core update fees, changes to OIS rewards, and the scheduled retirement of some push feeds show that products and incentives evolve. Publisher counts, supported chains, fees, and latency guarantees should be dated and sourced. Compare specific oracle products: for example, Chainlink Data Streams is also pull-based, so the broad claim that Chainlink is always push and Pyth is always pull is inaccurate.
Social Media Sentiment
A small sample of Reddit discussions reviewed for this page covered Pyth’s Cardano integration and a DAO economics update. Some commenters welcomed new integrations as an infrastructure improvement, while others questioned reliability, market share, and the metrics used to compare oracle networks. The October 8 discussion about routing DAO revenue to the PYTH Reserve was more positive about the transparency of the mechanism, but also raised questions about execution limits. This is a small, self-selected sample, not a measure of overall community opinion or evidence that technical claims are correct.
Last updated: 2026-10
Related Terms
- Oracle — software that supplies external data to a blockchain.
- Price Oracle — an oracle that provides price data.
- Oracle Manipulation — attempts to influence the data or results an oracle supplies.
See Also
- Chainlink Whitepaper — the design of Chainlink’s oracle network and data delivery model.
- Wormhole Whitepaper — cross-chain messaging infrastructure used in Pyth v2.0’s historical design.
- Band Protocol Whitepaper — another cross-chain oracle approach.
Sources
- Pyth Network Whitepaper v2.0, Pyth Data Association (September 28, 2023). Primary source for the Pythnet design, publisher and consumer model, fees described in v2.0, and that version’s aggregation algorithm.
- DAO proposal OP-PIP-100: Pyth Core migration to Pyth Pro. Explains the approved migration and plan to retire Pythnet.
- Pyth Core upgrade documentation. Describes the five routers, 3-of-5 signature threshold, and August 2026 infrastructure update.
- Pyth Pro architecture documentation and Pyth Pro price and confidence guide. Explain the current service, median aggregation, confidence values, and EMA.
- Pyth Core fee documentation and Oracle Integrity Staking documentation. Distinguish current Core fees from OIS reward and staking status.
- Notice on EVM Push Feed deprecations scheduled for October 15, 2026. Lists the feeds and chains affected.
- Chainlink Data Streams documentation. Shows why push and pull comparisons need to be made per product.
- Reddit discussion of Pyth’s Cardano integration and Reddit discussion of DAO revenue and the PYTH Reserve. Limited examples for the sentiment section, not representative surveys.
Last updated: 2026-10-10