Hyperion is our in-memory account-indexing engine, fed by Yellowstone gRPC. Account queries the rest of the market gates, rejects, or runs in seconds. Hyperion serves them in under a millisecond.
Included on every paid SVS RPC tier.
Measured against the same JSON-RPC method calls on a leading provider, on Solana mainnet, April 2026. Same payload, same network conditions, same account-set inputs. Reproducible. Full methodology in the docs.
| Method | Hyperion | Leading provider | Speedup |
|---|---|---|---|
| getTokenAccountsByOwner | 0.5 ms | 137 ms | 251× |
| getTokenAccountBalance | 0.3 ms | 174 ms | 521× |
| getSupply | 0.4 ms | 6,120 ms | 16,106× |
| getTokenLargestAccounts (USDC) | 2.7 s | rejected | n/a |
| getLargestAccounts | 0.4 ms | "contact support" | n/a |
| getTokenAccountsByDelegate | available | secondary index unavailable | n/a |
| getProgramAccounts | sub-10 ms | seconds to timeout | n/a |
Yellowstone Geyser streams every account write off our co-located validators. Hyperion fans those writes into purpose-built in-memory indexes, one per program. JSON-RPC requests hit the matching index instead of scanning a hot Postgres table or a remote gPA call.
Anyone whose latency budget gets eaten by account-query calls. Sub-millisecond reads compound across cycles: agents, arb bots, and dashboards see the difference immediately.
Hyperion is included on every paid RPC tier. Same URL, same key, same JSON-RPC spec.