Published measurements, their baselines and their limits. A fast connection is useful evidence. It is not a privacy guarantee.
No measured-speed ranking or speed metascore. These are 20 manually checked snapshots from 3 publishers, not our measurements. All are excluded from scoring until the missing test context is resolved. The three tables use different test methods and must be read separately.
RTINGS observations captured for 5 of 10 listed services; Top10VPN local-download observations captured for 10. A missing source observation is an evidence gap, not a failed speed test.
RTINGS: rolling multi-route averages.
RTINGS describes its method: repeated tests from a VPS to US East Coast, US West Coast, London and Sydney destinations, summarized over a rolling three-month window. These averages answer a different question from a nearby-server peak test. Per-provider platform and protocol, exact window bounds and sample counts have not been captured here.
Bar scale: 0–247 Mbps. Alphabetical order. Source snapshots checked 2026-11-06; exact test windows not captured. Provisional or held; excluded from scoring. This cohort has its own scale and method; compare within it only. Full baselines and limitations are in the table below.
Scroll horizontally to compare download, upload and limitations. Alphabetical provider order; rounded source values retained.
RTINGS paid-service rolling speed snapshots; provisional and held, excluded from scoring
Provisional snapshotPublisher says performance testing is ongoing. Exact rolling-window dates, per-provider protocol and sample count are not captured.Checked 2026-11-06 · testbench 0.11
A writing update, testbench conversion or kill-switch retest does not establish a new speed-test date. Read the source graphs for timing and variability. Download the observation fields.
Top10VPN: nearby-server download results.
Top10VPN describes its method. These paid-service values come from the local-download summary in each review. Its fixed 100 Mbps baseline differs from RTINGS’ setup. Do not compare the tables as a fastest-VPN ranking. Upload and latency values have not been captured for this cohort.
Bar scale: 0–100 Mbps. Alphabetical order. Source snapshots checked 2026-11-06; exact test windows not captured. Provisional or held; excluded from scoring. This cohort has its own scale and method; compare within it only. Full baselines and limitations are in the table below.
Scroll horizontally for baseline and test-context fields.
Top10VPN paid-service local download snapshots; provisional, excluded from scoring
100 MbpsPublisher’s fixed baseline; matched per-run baseline not captured
LightwayExact test dates, sample count, platform and endpoints not captured. Provisional; excluded from scoring.Article updated 2026-09-22; checked 2026-11-06
100 MbpsPublisher’s fixed baseline; matched per-run baseline not captured
NordLynxExact test dates, sample count, platform and endpoints not captured. Provisional; excluded from scoring.Article updated 2026-10-22; checked 2026-11-06
100 MbpsPublisher’s fixed baseline; matched per-run baseline not captured
WireGuardExact test dates, sample count, platform and endpoints not captured. Provisional; excluded from scoring.Article updated 2026-10-14; checked 2026-11-06
100 MbpsPublisher’s fixed baseline; matched per-run baseline not captured
Protocol not capturedExact test dates, sample count, platform and endpoints not captured. Provisional; excluded from scoring.Article updated 2026-10-14; checked 2026-11-06
100 MbpsPublisher’s fixed baseline; matched per-run baseline not captured
Protocol not capturedExact test dates, sample count, platform and endpoints not captured. Provisional; excluded from scoring.Article updated 2026-09-14; checked 2026-11-06
100 MbpsPublisher’s fixed baseline; matched per-run baseline not captured
Protocol not capturedExact test dates, sample count, platform and endpoints not captured. Provisional; excluded from scoring.Article updated 2026-10-14; checked 2026-11-06
100 MbpsPublisher’s fixed baseline; matched per-run baseline not captured
Protocol not capturedExact test dates, sample count, platform and endpoints not captured. Provisional; excluded from scoring.Article updated 2026-07-10; checked 2026-11-06
100 MbpsPublisher’s fixed baseline; matched per-run baseline not captured
Protocol not capturedExact test dates, sample count, platform and endpoints not captured. Provisional; excluded from scoring.Article updated 2024-11-25; checked 2026-11-06
100 MbpsPublisher’s fixed baseline; matched per-run baseline not captured
Protocol not capturedExact test dates, sample count, platform and endpoints not captured. Provisional; excluded from scoring.Article updated 2026-08-27; checked 2026-11-06
100 MbpsPublisher’s fixed baseline; matched per-run baseline not captured
Protocol not capturedExact test dates, sample count, platform and endpoints not captured. Provisional; excluded from scoring.Article updated 2026-08-19; checked 2026-11-06
PCWorld: percentage of baseline speed retained.
PCWorld describes its method: tests across six countries on multiple days, against a connection without the VPN. Higher percentages mean more of that baseline was retained. These are rounded article summaries, with qualifiers preserved. Absolute baseline Mbps, raw samples and actual test dates have not been captured; these historical observations cannot establish current performance or be compared directly with the other tables.
Bar scale: 0–100%. Alphabetical order. Source snapshots checked 2026-11-06; exact test windows not captured. Provisional or held; excluded from scoring. This cohort has its own scale and method; compare within it only. Full baselines and limitations are in the table below.
Scroll horizontally for upload results and dated evidence limits.
PCWorld percentage of baseline speed retained; provisional and excluded from scoring
Protocol not capturedPublished 2026-01-22; metadata updated 2026-01-22. These are not test dates.Actual test window, raw baseline speeds, sample count and exact endpoints not captured. Historical article evidence, not a current retest. Checked 2026-11-06. Excluded from scoring.
Protocol not capturedPublished 2026-03-17; metadata updated 2026-03-17. These are not test dates.Actual test window, raw baseline speeds, sample count and exact endpoints not captured. Historical article evidence, not a current retest. Checked 2026-11-06. Excluded from scoring.
WireGuardPublished 2025-05-19; metadata updated 2025-05-19. These are not test dates.Actual test window, raw baseline speeds, sample count and exact endpoints not captured. Historical article evidence, not a current retest. Checked 2026-11-06. Excluded from scoring.
Protocol not capturedPublished 2024-09-27; metadata updated 2025-06-18. These are not test dates.Actual test window, raw baseline speeds, sample count and exact endpoints not captured. Historical article evidence, not a current retest. Checked 2026-11-06. Excluded from scoring.
Protocol not capturedPublished 2024-03-25; metadata updated 2025-06-19. These are not test dates.Actual test window, raw baseline speeds, sample count and exact endpoints not captured. Historical article evidence, not a current retest. Checked 2026-11-06. Excluded from scoring.
Commissioned performance and security-feature comparison. The report describes differing test timing/reference conditions for the commissioning vendor, so a shared report is not proof of a matched cohort. Selected listed providers are covered; chart values are not transcribed.
