GetBlock RPC competitor monitor — 2026-10-28
Criterion of the day: Benchmark methodology and independent proof: transparent latency tests, third-party benchmarks and reproducible methodology (benchmark-methodology-independent-proof).
Executive summary
Today’s substantive change is QuickNode’s public live benchmark surface, not the numerous homepage/blog hash changes. Direct inspection of QuickNode Benchmarks verified both comparative result views and a meaningful methodology disclosure. It did not verify independent ownership or a complete reproducibility bundle. The cumulative topic matrix was therefore updated from 2/15 weakly evidenced providers to 1/15 conservatively evidenced provider: QuickNode at 2/4 dimensions.
No changed hash, navigation refresh, blog listing update or chain-directory update was treated as a market change by itself.
Decision-relevant findings
1. QuickNode: benchmark and methodology are now directly evidenced
- Live benchmarks expose chain/region/window comparisons and report P50, average, P95, success rate and data-freshness/block-lag concepts.
- The page describes 15 synthetic hosts across AWS, Google Cloud and Oracle Cloud in five regions (N. Virginia, California, Frankfurt, Tokyo and Singapore).
- It states that probes use fresh connections, identical payload/chain/region/schedule, record every attempt, retain timeout/rate-limit/failure outcomes, do not silently retry, and calculate latency from successful responses.
- Classification: Published benchmark results = publicly evidenced; transparent methodology = publicly evidenced.
- Limitation: this remains a vendor-run benchmark. Today’s pass did not verify downloadable raw observations, probe code, endpoint-selection policy or a rerun package. Independent proof and reproducibility remain not verified.
- The monitored home and enterprise pages repeat “2.5x faster” positioning, but those claims were not used alone; the benchmark page is the evidence anchor.
2. Weak prior dRPC classification removed
- The prior topic brief credited dRPC with a published benchmark based on dRPC’s homepage.
- Today’s evidence shows generic “low latency” positioning, not comparative results with a disclosed test frame.
- The dRPC row was changed from 1/4 to 0/4. This is a quality correction, not evidence that dRPC performs poorly.
3. Lava, Chainstack and headline latency claims remain unverified as benchmarks
- Lava says provider performance is “benchmarked and auditable,” but the monitored material did not expose a result set or enough method detail; no positive benchmark cell was assigned.
- Chainstack links a “Performance Dashboard,” but the captured monitored pages did not expose comparable results or methodology; it becomes a research priority, not a checkmark.
- Ankr (“56 ms avg. response time”) and NOWNodes (response-time headlines) provide first-party metrics without a verified benchmark method in today’s evidence. They remain unverified.
Provider-by-provider evidence log
| Provider | Today’s benchmark-methodology assessment | Direct source / limitation |
|---|---|---|
| QuickNode | Results and methodology verified; independent proof and reproducible artifacts not verified | Benchmarks, home, enterprise |
| Alchemy | No benchmark-methodology evidence verified in monitored pages | Home, pricing, enterprise; docs-chains and customer-stories monitored URLs returned 404 |
| Infura | No benchmark-methodology evidence verified | Home, pricing, docs; blog URL redirected to MetaMask developer news |
| Chainstack | Performance Dashboard link observed, but results/method not captured and therefore not credited | Home, docs |
| Ankr | 56 ms average headline is not sufficient methodology evidence | Home, enterprise |
| Moralis | No benchmark-methodology evidence verified | Home, docs |
| NOWNodes | Homepage response-time figures lack verified test design | Home; monitored Postman docs URL returned 404 |
| dRPC | Prior benchmark credit removed; generic low-latency positioning only | Home, docs |
| Blast API | Not applicable to active procurement; service is deprecated | Deprecation notice; docs redirect to Alchemy |
| Blockdaemon | No benchmark-methodology evidence verified | Home, docs; monitored RPC and pricing URLs returned 404 |
| Tatum | No benchmark-methodology evidence verified | Home, docs; monitored supported-blockchains URL returned 404 |
| Pocket Network / Grove | Former Grove Portal is winding down; current Pocket surface was not comparable benchmark evidence | Grove farewell, Pocket Network; Grove docs DNS failed |
| LlamaRPC | Not assessable from monitored public surface | Home/chains returned 526; docs/status DNS failed |
| Allnodes | Not assessable from automated monitored surface | All five monitored URLs returned 403; absence of evidence is not a product claim |
| Lava Network | “Benchmarked and auditable” statement observed, but no verified public results/methodology | Home, docs; developers/network URLs returned 404 |
Change-noise assessment
- QuickNode home, enterprise, blog and chains: changed snapshots include navigation, chain count, content listing and positioning updates. Only the direct benchmark surface created a matrix change.
- Alchemy, Infura, Chainstack, Ankr and NOWNodes blogs: changed listings are content churn, not benchmark proof.
- dRPC chain list: changed chain-directory content is unrelated to the criterion.
- 404/403/526/DNS failures: recorded as collection limitations. They are never converted into “No capability” claims.
- CompareNodes reference: the available SOC/ISO directory was reachable but is scoped to compliance. It was not used as independent latency proof for today’s criterion.
Topic brief changes
- QuickNode:
Published benchmarkretained and strengthened with the direct live benchmark URL;Methodologymoved from unverified to publicly evidenced; independent and reproducibility cells remain unverified. - dRPC:
Published benchmarkmoved from evidenced to not verified, because the prior home-page source did not satisfy the comparison contract. - Lava Network: added explicit vendor “benchmarked and auditable” observation, but kept all positive cells unverified pending result data and methodology.
- Chainstack, Ankr and NOWNodes: added precise caveats explaining why dashboard links or headline latency figures do not qualify.
- Blast API and Pocket Network/Grove: updated lifecycle caveats; LlamaRPC and Allnodes: updated collection limitations.
- Cross-provider totals, buyer guidance, direct sources and research priorities were recalculated.
GetBlock implications
- Sales: do not counter “2.5x faster” with an unsupported number. Ask whether the buyer accepts a vendor-run benchmark and offer a controlled proof-of-concept using the buyer’s chain, methods, regions and endpoint tier.
- Product marketing: GetBlock can differentiate by publishing a versioned, failure-inclusive methodology with endpoint-tier rules, P50/P95/P99, freshness, sample counts and raw exports.
- Content: a neutral “How to evaluate RPC benchmarks” page can explain cold vs warm connections, retries, tail latency, stale reads and why raw data/independence matter—without claiming a winner.
Publication and automation status
- Pre-run encyclopedia publication reported success (
published_urls: 142), but publication must be rerun after this report and topic brief update. - Agentic routing paper updater reported
ok: true,paper_updated: true, with 57 evidence items. Its update referenced QuickNode latency/SLA evidence and flagged cost/SLA constraints, method-specific routing, metrics and incident simulation for review. - Living paper target: Agentic Routing for RPC.
Sources checked
- QuickNode live benchmarks — direct inspection 2026-10-28.
- All provider home/pricing/enterprise/blog/docs URLs listed in the snapshot and normalized above — checked 2026-10-28.
- CompareNodes SOC/ISO directory — reachable 2026-10-28, but out of scope as benchmark proof.
> Volatile latency, uptime, chain-count and SLA claims must be re-verified before external sales use.