RPC provider performance on Arc: a regional latency and reliability benchmark (October 2026)

By Chainstack Labs Research Team · Research Note No. 049 · October 2026 edition · Published Oct 8, 2026 · Last updated Oct 8, 2026 · Node infrastructure

Data source: Chainstack Compare, backed by a public Grafana dashboard. Seven-day rolling window across DE, US, and SG, snapshot captured Oct 8, 2026.

Disclosure. Chainstack operates the monitoring service, the dashboard, and the scoring formula behind this report, and Chainstack is one of the four providers ranked. Section 2 documents the pipeline so readers can check the work, and the collection code is open source.

---

Abstract

Arc is a young enough network that public RPC comparisons barely exist, and the ones that do rarely account for region. This report ranks Chainstack, Quicknode, Alchemy, and dRPC on Arc RPC latency and uptime over a 7-day window ending Oct 8, 2026, using the scoring method behind Chainstack Compare: response time and success rate are combined per region, and the three regions (DE, US, and SG) are averaged equally. Chainstack ranks first with the flattest regional profile of the four providers.

Keywords: Arc RPC provider, Arc RPC latency, RPC benchmark, regional latency, availability, composite scoring, tail latency, methodology

Key findings

• Chainstack ranks first with a score of 0.059, 100.00% availability, and the lowest P95 (59 ms) and P99 (99 ms) in the set.
• Its regional profile is the flattest. P95 stays between 48 ms and 80 ms in every region, against 21–175 ms for Quicknode and 21–454 ms for Alchemy.
• Alchemy has the best DE (21 ms) and US (37 ms) cells but ranks third. Its SG P95 of 454 ms, up from 382 ms in September, is averaged in at equal weight.
• dRPC has the lowest median (14 ms), but a P95 of 180 ms and a P99 of 412 ms.
• Rank order matches September. Alchemy (0.173) and dRPC (0.181) are now nearly tied.

---

Contents

1. Introduction
2. Methodology and scoring
3. Providers and regions
4. Results
5. Regional variance and outliers
6. Changes since September
7. Interpretation notes and known confounders
8. Conclusion
9. Frequently asked questions
10. Appendix A: measurement parameters

---

1. Introduction

Most public RPC comparisons report one global latency number per provider, which hides the failure mode that hurts integrators most: a provider that is fast where it has infrastructure and slow everywhere else. A trading bot running in Singapore does not benefit from a figure measured in Germany. This report ranks four providers on Arc using a score built to surface that gap, and publishes the regional detail behind the ranking.

Arc is a younger network than Ethereum or BNB Chain, so provider footprints and routing are still settling. Expect more movement between editions than on older chains.

2. Methodology and scoring

2.1 Monitoring infrastructure

A Python service running on Vercel serverless functions sends RPC requests to each monitored provider every 1–3 minutes from three regions and records each request’s response time and outcome. The results feed a public Grafana dashboard, which also powers Chainstack Compare. Regions are labeled by country code: DE, US, and SG. Because probes originate from cloud regions, the benchmark measures latency between each provider and those regions, not the latency an end user on a home or mobile connection would see.

2.2 What is requested and what counts as a failure

All requests use the latest block, except methods marked with an hourglass (⌛) on the dashboard. These “archive” methods query historical blocks one to two days old, which exercises a provider’s historical-state path rather than its tip-of-chain path. A response that takes longer than 55 seconds is treated as a failure, not as a very slow success, and is excluded from the latency charts. Per-method breakdowns are available in Grafana; this report uses the summary view.

2.3 Plan monitored

Unless a different plan is specified, each provider is monitored on its first paid RPC plan, named Growth as a rule. The results describe each provider’s entry paid tier. Dedicated, enterprise, and other higher tiers are not covered, and rate limits or routing that differ by plan may affect both latency and success rate.

2.4 Scoring formula

The dashboard ranks providers on speed (response time) and reliability (success rate). Lower is better.

Score = 1 / ((1 / ResponseTime) × (SuccessRate³))      = ResponseTime / SuccessRate³

The score is built in four steps:

1. Latency term. For each region, take each method’s p95 and combine them as an arithmetic mean with equal weight per method. eth_subscribe is excluded from this average because its latency reflects chain block cadence, not provider speed.
2. Success-rate term. For each region, compute the success rate using sample counts as weights. eth_subscribe still counts toward this term.
3. Per-region score. Divide the latency term by the success rate cubed. Response time is in seconds, so a latency term of 0.059 means 59 ms.
4. Final score. Take the arithmetic mean of the regional scores. Every region carries equal weight, so one slow region cannot be hidden by fast ones.

2.5 How much the success-rate cube matters

Cubing the success rate makes small reliability drops visible, but at the availability levels measured on Arc the effect is small.

Success rate	Score multiplier
100.00%	1.000×
99.90%	1.003×
99.50%	1.015×
99.00%	1.031×
97.00%	1.096×
95.00%	1.166×

The lowest availability on Arc is Alchemy’s 99.82%, a multiplier of about 1.005×. Alchemy’s third-place score is driven by its SG latency, not its availability. Success rate becomes a dominant term only well below 99%.

2.6 Reading the dashboard columns on Arc

The P95 column is the mean of the three regional P95 values. Chainstack (49, 80, and 48 ms) averages 59.0 ms against 59 shown; Quicknode (21, 91, 175) averages 95.7 against 96; Alchemy (21, 37, 454) averages 170.7 against 171; dRPC (92, 254, 194) averages 180.0 against 180.

On Arc, the score also tracks that column closely: every score falls within 1% of mean P95 divided by availability cubed. On Ethereum, BNB Chain, and Hyperliquid the two diverge by up to about 15%, so Arc is the chain where the percentile columns are the most reliable stand-in for the score. Rank order is identical under both measures on every chain.

2.7 Time window, limits, and independence

The dashboard is live, refreshes every 3 minutes, and offers 24-hour and 7-day views. Every figure here comes from the 7-day view captured on Oct 8, 2026. The benchmark does not measure end-user latency, throughput under load, response correctness, cost, or tiers above the entry paid plan. Chainstack chose the formula (p95, success rate cubed, equal regional weights) and operates the probes, so a different formula could reorder close pairs.

3. Providers and regions

Four providers with public Arc RPC endpoints are monitored: Chainstack, Quicknode, Alchemy, and dRPC. Each is probed concurrently from DE, US, and SG over the same rolling 7-day window, on its first paid plan. To build on Arc with Chainstack, see the Chainstack Arc page.

4. Results

Table 1. Arc, 7-day window, Oct 8, 2026. Chainstack leads on score, P95, and P99; Chainstack and Quicknode both record 100.00% availability; dRPC has the lowest P50.

Provider	Score	Availability	P50	P95	P99	DE	US	SG
Chainstack	0.059	100.00%	19ms	59ms	99ms	49ms	80ms	48ms
Quicknode	0.096	100.00%	82ms	96ms	113ms	21ms	91ms	175ms
Alchemy	0.173	99.82%	49ms	171ms	224ms	21ms	37ms	454ms
dRPC	0.181	99.99%	14ms	180ms	412ms	92ms	254ms	194ms

Quicknode’s score is 63% higher than Chainstack’s. Alchemy and dRPC are close to each other (0.173 and 0.181) and about three times Chainstack’s score.

5. Regional variance and outliers

Alchemy shows the sharpest cliff. Its DE (21 ms) and US (37 ms) figures are the best or joint-best in the set, and its SG (454 ms) is the largest cell outside dRPC’s Ethereum results in this series. The 21.6× gap between its best and worst region is why it ranks third despite leading two regions. Equal regional weighting is what produces that result.

Quicknode is fast in DE and slow in SG. It ties Alchemy at 21 ms in DE, then rises to 91 ms in US and 175 ms in SG, an 8.3× spread. Its P50 (82 ms) is the highest in the set, but its P95 (96 ms) and P99 (113 ms) are tightly clustered, which suggests consistent behavior rather than occasional spikes.

dRPC has the lowest median and a long tail. Its P50 of 14 ms is the fastest on Arc, but its P95 (180 ms) and P99 (412 ms) are far higher, and its US cell (254 ms) is the weakest outside Alchemy’s SG.

Chainstack has the flattest profile. Its regional range is 48–80 ms (1.7×), against 21–175 ms for Quicknode (8.3×), 21–454 ms for Alchemy (21.6×), and 92–254 ms for dRPC (2.8×). It leads only one region outright (SG, 48 ms), concedes DE to Quicknode and Alchemy (21 ms), and concedes US to Alchemy (37 ms). The score rewards that consistency because it averages regions equally.

Table 2. Lowest P95 by region on Arc.

Region	Lowest P95
DE	Quicknode and Alchemy (21 ms, tie)
US	Alchemy (37 ms)
SG	Chainstack (48 ms)

6. Changes since September

The September 2026 edition of this report covered the same four providers on Arc. Table 3 compares its figures with this month’s.

Table 3. Arc, September snapshot (Sep 21, 2026) to October snapshot (Oct 8, 2026). Lower is better.

Provider	Score	P95 (global)	Availability	Worst region P95
Chainstack	0.063 → 0.059	62 → 59 ms	99.98% → 100.00%	US 79 → US 80 ms
Quicknode	0.101 → 0.096	101 → 96 ms	99.97% → 100.00%	SG 177 → SG 175 ms
Alchemy	0.149 → 0.173	149 → 171 ms	99.98% → 99.82%	SG 382 → SG 454 ms
dRPC	0.224 → 0.181	238 → 180 ms	99.88% → 99.99%	US 311 → US 254 ms

Rank order held. Chainstack, Quicknode, Alchemy, and dRPC finished in the same order as in September, though Alchemy and dRPC are now nearly tied (0.173 and 0.181).

Two regional weaknesses persisted. Alchemy’s SG latency worsened from 382 ms to 454 ms, and Quicknode’s SG latency was essentially unchanged (177 ms to 175 ms). Two consecutive windows do not prove a structural cause, but they make a one-off less likely, and both providers’ DE and US cells remain strong.

dRPC improved on every measure (score, P95, and availability from 99.88% to 99.99%), though its US cell is still 254 ms. Each snapshot is one 7-day window, so this shows two points in time, not a trend.

7. Interpretation notes and known confounders

Entry paid plan only. Higher tiers, including dedicated infrastructure placed near a weak region such as SG, could change a provider’s regional profile and are not tested.

Probes are synthetic and regional. Requests come from serverless functions in three cloud regions. A team whose users sit elsewhere should treat the nearest region as a proxy and verify with its own probes.

Arc is an early-stage network. Its traffic profile and provider footprints are still maturing. Absolute figures can shift independent of any change in provider quality.

Method mix is summarized, not shown. Headline numbers average across methods with equal weights. Methods differ widely in cost, so a workload concentrated in one heavy method may rank providers differently. Per-method values are in Grafana.

Single window. Both the October figures and the September comparison rest on a single 7-day window each.

8. Conclusion

On Arc, ranking by regional consistency instead of peak speed changes the picture. Quicknode and Alchemy both post the best figure in at least one region, but neither holds it across all three, and the score is built to penalize that gap. Chainstack leads with a profile that never rises above 80 ms. Teams choosing an RPC provider for Arc should decide whether peak regional speed or global consistency matches where their users are: a DE- or US-only workload may reasonably prefer Alchemy or Quicknode, while anything touching SG should look closely at Table 1 before choosing.

9. Frequently asked questions

Which RPC provider has the lowest latency on Arc?

Chainstack, on the 7-day window ending Oct 8, 2026. It scores 0.059 with a global P95 of 59 ms, against 96 ms for Quicknode, 171 ms for Alchemy, and 180 ms for dRPC.

Which provider is fastest in DE and US on Arc?

Alchemy leads both regions, with a DE P95 of 21 ms (tied with Quicknode) and a US P95 of 37 ms. Chainstack posts 49 ms in DE and 80 ms in US.

Why does Alchemy rank third on Arc despite the fastest DE and US figures?

Its SG P95 is 454 ms. The score averages the three regional scores with equal weight, so one very slow region offsets two excellent ones (Section 5).

Appendix A: measurement parameters

Parameter	Value
Chain	Arc
Providers	Chainstack, Quicknode, Alchemy, dRPC
Plan monitored	First paid RPC plan per provider (Growth as a rule)
Regions	DE, US, SG
Probe infrastructure	Python service on Vercel serverless functions
Probe cadence	Every 1–3 minutes per region
Request target	Latest block; archive methods (⌛) query blocks 1–2 days old
Failure threshold	Responses over 55 seconds count as failures
Latency term	Mean of per-method p95, equal method weights, eth_subscribe excluded
Success-rate term	Sample-count weighted, eth_subscribe included
Score formula	ResponseTime / SuccessRate³ per region, then arithmetic mean across regions
Time window	Rolling 7 days; dashboard refreshes every 3 minutes
Snapshot date	Oct 8, 2026
Source	Chainstack Compare, public Grafana dashboard
Collection code	compare-dashboard-functions