Hyperliquid's HIP-4, launched in early May 2026, brings binary outcome markets to its on-chain central limit order book. The contracts are fully collateralized, settle at 1 or 0, and trade alongside spot and perpetuals using the same account. It's a direct shot at Polymarket — but the two platforms diverge on the most critical question: who gets to call the winner.
How HIP-4 works
HIP-4 uses a merged order book, where a buy order for Yes at price p is identical to a sell order for No at 1-p. That pools liquidity on both sides, a design Polymarket also uses. The difference is where the market lives. HIP-4 is a primitive baked into HyperCore, Hyperliquid's base layer, not a separate app. Anyone can create a market via the builder-deployed framework (HIP-3). The first markets were recurring daily Bitcoin price binaries; later ones expanded to macro events like US inflation prints and Federal Reserve decisions. For now, fees are zero during a testing period.
Polymarket's setup
Polymarket runs on Polygon, uses Gnosis Conditional Tokens, settles in USDC, and outsources resolution to UMA or other oracles. Its monthly volume peaked near $5 billion in March 2026 per DeFiLlama (Polymarket's own numbers are about double that). That's a lot of action. But the architecture is fundamentally different from Hyperliquid's — and that matters for how disputes get settled.
The real battleground: resolution
HIP-4 uses validator-based outcome resolution. Hyperliquid's validators decide the result of each market. Polymarket relies on UMA's optimistic oracle or other third-party oracles. The article argues the real competition isn't over fees — it's over who resolves reality. Validator resolution is faster and more integrated, but it concentrates power in the hands of Hyperliquid's validator set. Oracle resolution is more decentralized but slower and can be gamed. Neither is perfect. The question is which trade-off users trust more.
The zero-fee period won't last forever. Once it ends, the fee structure will test whether liquidity migrates from Polymarket or stays put. Meanwhile, HIP-4's first macro-event markets are already live. The next concrete step: more complex event types, like multi-outcome or conditional binaries, could push validator resolution to its limits. That's the unresolved question — can a small validator set handle the nuance of a thousand bespoke markets without centralizing truth?
