HyperCore market data

Elysium serves live HyperCore state to every contract through a chain-native precompile. Orderbooks, prices, balances, and positions are simply there to read, for ordinary call gas, with no oracle contracts, relay hops, or per-update push costs.

The read precompile ships in an ArbOS network upgrade about four weeks after mainnet. Elysium launches on stock ArbOS without it.

What you can read

• Orderbooks at a chosen granularity, from best bid and offer up to full depth.
• Prices: oracle, mark, and spot.
• Accounts: balances and positions.
• Computed aggregates: depth within ±bps of mid, flow imbalance, and expected slippage for a given size.

Data is available at HyperCore block granularity (≈70 ms), with ≈150–350 ms freshness inside the EVM.

How it works

Elysium runs its own Hyperliquid node, co-located with the chain, and serves HyperCore state from inside the execution layer. Nothing is pushed on-chain ahead of time. A contract queries the precompile and pays gas in proportion to the data it asks for, so a full-depth book read costs more than a best-bid-and-offer read, and neither costs a storage write.

Freshness and auditability

Every response is stamped with the HyperCore block height and timestamp it was derived from. Contracts should enforce their own freshness bounds against those values.

Each read also commits to the exact bytes it serves. This keeps reads deterministic when the chain is replayed and makes every response auditable against archived HyperCore data. If the data path falls behind, reads fail visibly stale rather than silently wrong.

The data is attested by the sequencer, not proven by HyperCore. HyperCore does not publish state roots or state proofs today, so reads anchor to Elysium's archived copy of HyperCore inputs. The long-term plan is fraud-provable attestation against that archive as permissionless proving matures.

Compared with HyperEVM and oracle networks