Network

Solana

SOL · 420 ms block time · 3 providers, 5 endpoints

Just tell meWhich Solana provider leadsThe ranking below, read as an answer — with the figures behind it and what it refuses to claim.Before you payDo you need a dedicated node?What the free endpoints actually manage, and how often they refuse. Often the answer is no.

Who is producing blocks

leader schedule · epoch 1052

Solana fixes its leader schedule for a whole epoch before the epoch begins and publishes it, so who produces each of the next 432,000 slots is a known fact rather than a forecast. No other chain measured here publishes one. The map beside it marks where the validator taking its turn was last observed on the network.

Block production

epoch 1052 · 35.0% through

slot 151,283 of 432,000 in this epoch · absolute 454,615,283

Leading now

9rkJMARqK6VBkcxGfKBAwnA44gPAfGxPbPsfsggFNDSQ

4,706,531.64 SOL staked

Last observed

—

projected since

Next up

+0.4s8BwBBt23a6v8HgNsFybpLFiBKLV8vWko2Vft3D2DxGvR4 slots245,055.89 SOL
+2.0s7RtC1QgiNVLANGRTEKxASz4bRiMYyf2hUVLs1XGuJ6Gr4 slots282,052.99 SOL
+3.6sDF1owXYZ1fk5vWyHJ8s1cJeAozgkqsi1JUVqvktqrpwd4 slots389,102.30 SOL
+5.2sHz5aLvpKScNWoe9YZWxBLrQA3qzHJivBGtfciMekk8m54 slots3,935,225.66 SOL
+6.8sJUPiTERrZqgf1jUyR7dSkhMx4Kn2qJyekWsg3LT1h4b4 slots11,075,221.92 SOL

The leader schedule and the observed slot are measured; the current slot is projected forward from the last sample at the chain's target slot time. Solana does not produce slots on a metronome, so treat the projection as an estimate and observed_at as the fact.

Producing from

19,801 located · 50 countries

Current leader

9rkJMARqK6VBkcxGfKBAwnA44gPAfGxPbPsfsggFNDSQ

Los Angeles, US · AS396356 Latitude.sh

Solana slots advance roughly every 400ms, with the observed rate slightly slower because of skipped slots. Commitment levels (processed, confirmed, finalized) mean two healthy providers can legitimately report different heads, so every head is observed at a fixed commitment and compared against the cohort rather than against any single provider.

Provider ranking

30d window · 30.0 days observed · 4 vantage points · generated 2026-10-08 17:45:14 UTC

Excellent / Good— Fast, available and correct all window
Fair / Degraded— Measurable failures or slow tails
Poor— Failing often enough to affect an application
Insufficient data— Too few samples to score at all

1 (1st)
Provider
PublicNode
Score
92.6
Status
Excellent
Availability
99.2%
p50 ms
22.8
p95 ms
50.8
Trend
steady
Correct
99.2%
Behind
9.52
Samples
1,324,561
2 (2nd)
Provider
Tatum
Score
83.4
Status
Good
Availability
99.2%
p50 ms
26.6
p95 ms
88.5
Trend
steady
Correct
99.3%
Behind
8.11
Samples
440,235
3 (3rd)
Provider
Solana Foundation
Score
69.9
Status
Degraded
Availability
99.8%
p50 ms
15.9
p95 ms
258.3
Trend
−77%
Correct
100.0%
Behind
0.03
Samples
1,008,320

Provider ranking by composite score
#	Provider	Score	Status	Availability	p50 ms	p95 ms	Trend	Correct	Behind	Samples
1 (1st)	PublicNode	92.6	Excellent	99.2%	22.8	50.8	steady	99.2%	9.52	1,324,561
2 (2nd)	Tatum	83.4	Good	99.2%	26.6	88.5	steady	99.3%	8.11	440,235
3 (3rd)	Solana Foundation	69.9	Degraded	99.8%	15.9	258.3	−77%	100.0%	0.03	1,008,320

Take it awayCSV JSONevery provider in this window, with each score component

publicnode — 0% of requests targeted methods this provider does not implement
tatum — 12% of requests were rate limited (49859 of 423014); excluded from availability
solana-foundation — 0% of requests targeted methods this provider does not implement

3 providers scored against each other. A cohort this small makes every rank a comparison between very few alternatives: first of 3 is a weaker claim than the same score on a crowded chain, and the score is relative to whoever is here.

The score weights availability at 40%, latency at 30%, correctness at 20% and head freshness at 10%. Correctness also acts as a ceiling: a provider that answers quickly and wrongly cannot be rated healthy on speed alone. Full methodology

Latency over time

30d window · top 3 by score

95th percentile, computed over successful requests only. A request that timed out is not a latency sample; it is counted against availability instead. Gaps are gaps — periods with no collection are not bridged.

Head freshness

30d window · generated 2026-10-08 17:45:14 UTC · cohort of 3

How far behind the chain each provider's data is. Latency cannot answer this: an endpoint can reply in 8 ms with a head from four minutes ago, and every speed measurement will still rate it excellent.

Providers ranked by distance behind the cohort head
#	Provider	Mean slots behind	≈ seconds	Worst slots	Compared
1 (1st)	solana-foundation	0.03	0.01	56	496,670
2 (2nd)	tatum	8.11	3.40	5,331	136,636
3 (3rd)	publicnode	9.52	4.00	1,651	712,601

Each observed head is compared against the cohort reference for its instant: the highest head any measured provider had reached, after discarding heads implausibly far above the cohort median and time-aligning samples taken moments apart. Lag is therefore relative to the fastest-updating provider in the set, not to an absolute truth. A cohort of fewer than two providers produces no freshness data at all.

From each region

The same Solana measurements, grouped by the vantage point they were taken from. A provider that answers in 40 ms from one region can take 220 ms from another, and the ranking above is an average across whichever regions are reporting.

Solana broken out by region
Region	Median	p95	p99	Success	Rate limited	Samples
Virginia	17.9 ms	98.4 ms	499 ms	95.1%	12,573	693,875
Amsterdam	23.1 ms	117 ms	684 ms	95.0%	12,531	693,189
Coventry	31.5 ms	146 ms	704 ms	94.8%	12,443	693,028
Frankfurt	16.0 ms	147 ms	656 ms	94.9%	12,312	693,024

Percentiles are over successful responses only. A request that timed out is not a latency sample; it is counted against availability instead.