Five venues run real markets on crypto prices—Kalshi, Polymarket, Predict.fun, Limitless, and Hyperliquid—and several describe themselves as “Chainlink-powered” in their own marketing. That phrase tells you less than it sounds like it does. Two platforms can both use it and mean structurally different things underneath.
Kalshi and Polymarket: an index versus an oracle
Kalshi’s short-dated crypto markets settle against the CF Benchmarks Real Time Index, an aggregate across multiple exchanges rather than any single feed. Polymarket’s 5-minute and 15-minute crypto markets settle through Chainlink Data Streams, while Pyth Network specifically handles its newer equities, commodities, and ETF markets.
An aggregated index and an oracle network are different designs solving the same problem differently. See the full infrastructure and resolution comparison for how and why they can diverge on a fast-moving print.
Predict.fun: also “Chainlink,” but not the same Chainlink
Predict.fun’s own announcement describes adopting Chainlink as core infrastructure for its 15-minute crypto markets, using the Chainlink Runtime Environment to automate market lifecycle and resolution and Chainlink’s DataLink product to publish the actual price data. Per that announcement, DataLink for Predict.fun accesses data from Binance’s order books specifically.
Chainlink’s standard methodology is normally a multi-exchange pipeline: raw data from centralized exchanges and on-chain DEXs is refined by independent aggregators, multiple node operators each take a median, and the network aggregates those results again into one report. Polymarket’s implementation reads as closer to that fuller version. Predict.fun’s, at least per its own description, traces back to one exchange. Both are accurately called “Chainlink-powered,” but they do not describe the same underlying data pipeline.
Limitless: two oracles, split by market duration
Limitless runs a dual-oracle setup rather than choosing one. Pyth Network is the primary price oracle for most of its data-feed markets. Chainlink Data Streams was added alongside it in June 2026, specifically for short-duration crypto markets, which by mid-2026 were running around 19,000 markets a week. Which oracle prices a Limitless market depends on that market’s duration, not one platform-wide choice.
Hyperliquid: not a resolution source, a blended median
Hyperliquid is not a prediction market and does not resolve a yes-or-no bet against an oracle. It is a perpetuals exchange, where the mark price matters for funding and liquidation. That mark price is the median of three inputs: a validator-computed oracle price plus a 150-second EMA adjustment against Hyperliquid’s own mid-price; the median of Hyperliquid’s best bid, best ask, and last trade; and a weighted median of perpetual mid-prices from Binance, OKX, Bybit, Gate.io, and MEXC.
The validator oracle is itself a weighted median of centralized-exchange spot prices, weighted by exchange liquidity and updated roughly every three seconds. Funding settles hourly from a premium index—the running gap between that blended mark price and the external spot oracle—sampled every five seconds.
The design logic is explicit: no single input, including Hyperliquid’s own order book, should move the mark price alone. That is a different problem from deciding which source resolves a contract, and a different answer from anything the four prediction-market venues are doing.
What this actually means
“Chainlink-powered,” “oracle-based,” and “index-settled” are not interchangeable. None tells you enough on its own to assume two venues price the same asset the same way at the same moment. Check the actual mechanism, not just the vendor name attached to it. A single-exchange-rooted feed and a five-exchange blended median can both be accurate most of the time and still diverge exactly when it matters: during a fast move on one venue’s reference exchange that the others have not yet absorbed.
Historical trade and price data across Kalshi, Polymarket, Predict.fun, Limitless, Deribit, and Binance options is available in one schema through the API docs if you want to check divergence on a specific crypto market yourself.
FAQ
Do Polymarket and Predict.fun use the same Chainlink price feed for crypto markets?
Not established as identical. Both describe their crypto markets as Chainlink-powered, but per Predict.fun’s own announcement, its implementation sources from Binance’s order books specifically, which is narrower than Chainlink’s standard multi-exchange aggregation methodology.
Does Hyperliquid use an oracle to resolve markets?
Not in the same sense as a prediction market. Hyperliquid computes a blended mark price from three inputs: a validator-computed oracle, its own order book, and a weighted median of five external exchanges’ perpetual prices, specifically so no single source can move it alone.
Which of these five venues uses the broadest multi-exchange price aggregation?
Hyperliquid’s mark-price mechanism blends the most distinct sources: its own book, a validator oracle drawing on multiple centralized exchanges, and five named external exchanges. Among the prediction-market venues, Kalshi’s CF Benchmarks index and Chainlink’s standard methodology are both explicitly multi-exchange by design.
Why does Limitless use two different oracles instead of one?
Per its own rollout, Pyth remains the default for most markets, with Chainlink Data Streams added specifically for short-duration crypto markets, where update speed matters more. Which oracle applies depends on the specific market’s duration.
Is a single-exchange-sourced price feed necessarily worse than a multi-exchange one?
Not automatically. It can be faster and simpler, and work well most of the time. Its risk is concentrated in the moments a multi-exchange design is meant to protect against: a bad print, an outage, or unusual activity on the single reference exchange.