Performance claims, with evidence attached.

Concurrent EIP-1559 transactions used the same sender and nonce with matched fees. Gas limits identified the routes; successful onchain receipts determined the winners.

Results cover the listed endpoints from one Virginia client on November 7–8, 2026; other regions, time windows and fee policies may differ.

The same Ohio-2b client triggered four matched-fee requests per round. Receipts identified same-nonce winners; block height and transactionIndex determined distinct-nonce ordering.

Results cover the listed endpoints and one Ohio-2b test window. One Eira Nodes transaction had no receipt, leaving 99 comparable pairwise samples.

Feeds were matched by block height and compared at client-ready time. Transaction tests used same-nonce inclusion and distinct-nonce block height plus Transaction Index.

Results apply to the published endpoints, routes, sample windows and client implementation. The report identifies the Ohio transaction relay but does not publish the Feed client region.

Regional clients applied category-specific rules: first delivery for node, block, and shared mempool events; union coverage for mempool monitoring; then inclusion block and same-block index for paired transaction submissions.

Results are specific to the stated test periods, regions, network conditions, and decision rules. First-arrival lead percentiles are conditional winning margins, not end-to-end latency.

One client timestamped each structured event when it became application-ready, matched events by block height and assigned 0 ms to the earlier source.

This compares practical application-ready paths over different protocols, not protocol-neutral raw transport latency.

The same client triggered matched transactions with no send failures. The tests compared same-nonce winners, block height and transaction index.

The result measures the published Ohio test conditions and does not guarantee a fixed ordering position.

The first feed to deliver a matching block received 0 ms relative latency. Later arrival was measured from the timestamp difference.

Zero relative latency means first arrival in the comparison, not zero end-to-end delivery time.

Each regional client submitted identical-nonce pairs to BlockRazor and the official endpoint with matching parameters.

The result compares the stated regions and transaction configuration, not every production workload.

Regional clients compared transaction position and timestamp differences for matching blocks and FlashBlocks.

Submission and stream metrics describe different tests and should not be combined into one latency score.

The benchmark timed transaction construction, submission and landing confirmation with Shred parsing or Yellowstone gRPC.

Client and server results include different network and application layers and should be read separately.

Clients received both streams, deduplicated by slot and Shred index, then calculated first arrival and lead-time percentiles.

The test measures Shred delivery timing, not transaction execution or strategy outcomes.

The test held tip and priority fee constant, excluded Jito-landed transactions and compared which provider reached the Leader first.

The result describes the filtered SWQoS route under equal-fee conditions.

For matching transactions, regional clients measured the reception-time difference between BlockRazor and a regular node.

The documentation does not publish a sample count or test date, so the card does not infer either value.

Clients recorded reception timestamps for matching blocks from the regional relay and a regular node.

The documentation does not publish a sample count or test date.

The benchmark compared the same node with and without the BlockRazor Relay peer connection.

The report does not publish a sample count or test date.

Physically close clients compared first-seen transactions and blocks across equivalent service plans.

This is a historical 2024 benchmark. Product plans, networks and competitors may have changed.