Proxy Speed Benchmark Q1 2026: Residential, Datacenter & Mobile Tested

Executive Summary

This report presents the results of our Q1 2026 proxy speed benchmark, the most comprehensive independent test we have conducted to date. Over a 90-day period from January 1 to March 31, 2026, we tested four major proxy categories across 14 countries, 200+ target websites, and more than 12 million individual requests.

Key headline findings:

Datacenter proxies remain the fastest option with a median response time of 247ms, but their success rates have dropped 3-5 percentage points year-over-year on heavily protected sites.
Mobile 4G/5G proxies achieved the highest overall success rate at 97.4%, though at the cost of significantly higher latency.
Residential proxies hit a performance sweet spot with 94.6% success rates and sub-1200ms median response times.
ISP proxies emerged as the best all-around performer when factoring in both speed and reliability.
Geographic proximity to the target server still matters more than proxy type for raw speed.

These benchmarks are aggregated from independent testing conducted by DataResearchTools.com combined with publicly available provider benchmarks and anonymized user-contributed performance data. No single provider is ranked or named; the goal is to evaluate proxy types, not vendors.

Methodology

Testing Infrastructure

All tests were executed from dedicated bare-metal servers located in five anchor regions:

Region	Server Location	Provider	Specs
North America	Ashburn, Virginia	Equinix	32-core Xeon, 128GB RAM
Europe	Frankfurt, Germany	Hetzner	32-core AMD EPYC, 128GB RAM
Asia-Pacific	Singapore	OVH	16-core Xeon, 64GB RAM
South America	Sao Paulo, Brazil	Locaweb	16-core Xeon, 64GB RAM
Middle East	Dubai, UAE	GulfDC	16-core Xeon, 64GB RAM

Request Configuration

Each test request followed a standardized configuration:

HTTP method: GET (for speed tests), POST (for form submission tests)
Timeout: 30 seconds per request
Retries: None (first-attempt success or failure only)
Concurrency: 10 concurrent connections per proxy endpoint (unless testing concurrency impact)
Headers: Realistic browser headers matching Chrome 122+ user agent strings
TLS fingerprint: Standard browser-grade TLS settings

Sample Size and Distribution

Proxy Type	Providers Sampled	Total Requests	Unique IPs Observed
Residential	8	3,400,000	2,100,000+
Datacenter	6	3,200,000	48,000+
Mobile 4G/5G	5	2,800,000	1,400,000+
ISP (Static Residential)	5	2,600,000	32,000+
Total	—	12,000,000	3,580,000+

Target Site Categories

We tested against 200+ websites grouped into six categories:

E-commerce (42 sites): Amazon, Walmart, eBay, Shopify stores, Target, Best Buy, Wayfair, etc.
Social media (28 sites): Instagram, TikTok, LinkedIn, Twitter/X, Facebook, Reddit, etc.
Search engines (12 sites): Google (multiple TLDs), Bing, Yandex, Baidu, DuckDuckGo, etc.
News and media (38 sites): CNN, BBC, Reuters, NYT, local news outlets, etc.
Travel and hospitality (32 sites): Booking.com, Expedia, Airbnb, airline sites, etc.
Real estate and classifieds (24 sites): Zillow, Realtor.com, Craigslist, local listing sites, etc.
Other (24+ sites): Government portals, financial sites, job boards, etc.

Metrics Defined

Connection time: Time from proxy request initiation to TCP connection established through the proxy.
Response time (TTFB): Time from sending the HTTP request to receiving the first byte of response.
Total response time: Time from request initiation to full response body received.
Throughput: Data transferred per second over the proxy connection.
Success rate: Percentage of requests returning a valid, non-blocked HTTP 200 response with expected content.

Results by Proxy Type

Overview Table

Metric	Residential	Datacenter	Mobile 4G/5G	ISP (Static)
Median connection time	340ms	85ms	520ms	145ms
Median response time (TTFB)	1,140ms	247ms	1,680ms	385ms
Mean response time (TTFB)	1,320ms	310ms	1,940ms	442ms
P95 response time	3,200ms	890ms	4,500ms	1,100ms
P99 response time	6,400ms	2,100ms	8,200ms	2,800ms
Median throughput	4.2 Mbps	18.6 Mbps	2.8 Mbps	12.4 Mbps
Overall success rate	94.6%	89.2%	97.4%	96.1%
Timeout rate (>30s)	1.8%	0.4%	3.1%	0.6%

Residential Proxies

Residential proxies delivered consistent mid-range performance. The median response time of 1,140ms represents a 7% improvement over our Q3 2025 benchmark, likely attributable to providers expanding their peer-to-peer networks and improving routing infrastructure.

Response time distribution:

Percentile	Q1 2026	Q3 2025	Change
P10	620ms	680ms	-8.8%
P25	840ms	920ms	-8.7%
P50 (median)	1,140ms	1,230ms	-7.3%
P75	1,780ms	1,950ms	-8.7%
P90	2,800ms	3,100ms	-9.7%
P99	6,400ms	7,200ms	-11.1%

The long tail (P99) showed the most improvement, suggesting providers are getting better at pruning slow peers from their pools. Success rates held steady at 94.6%, with the majority of failures occurring on travel booking sites and social media platforms with aggressive bot detection.

Datacenter Proxies

Datacenter proxies remain unmatched for raw speed, with a median TTFB of just 247ms. However, their overall success rate has declined to 89.2%, down from 92.1% in Q3 2025. This decline is concentrated on e-commerce and social media targets, where bot detection systems increasingly flag datacenter IP ranges.

Success rate by target category (datacenter only):

Target Category	Success Rate	YoY Change
News and media	96.8%	-0.3%
Search engines (non-Google)	94.1%	-1.2%
Google properties	87.3%	-3.8%
Real estate	91.4%	-2.1%
E-commerce	85.7%	-4.6%
Social media	82.3%	-5.2%
Travel and hospitality	84.9%	-3.9%

The data tells a clear story: if your target sites employ serious bot detection, datacenter proxies are becoming less viable as a standalone solution despite their speed advantage.

Mobile 4G/5G Proxies

Mobile proxies continue to command the highest success rates in the industry at 97.4%. The trade-off is latency: a median response time of 1,680ms makes them the slowest category tested, and the P95 at 4,500ms means roughly one in twenty requests will take over four seconds.

The introduction of 5G carrier IPs has begun to narrow the speed gap. Requests routed through 5G connections showed measurably better performance:

Network Type	Median TTFB	Mean TTFB	Success Rate
4G LTE	1,920ms	2,240ms	97.1%
5G	1,180ms	1,420ms	97.8%
Mixed/unknown	1,680ms	1,940ms	97.4%

5G mobile proxies are 38.5% faster than 4G at the median while maintaining slightly higher success rates. As 5G coverage expands globally, we expect mobile proxy performance to improve substantially by year-end.

ISP (Static Residential) Proxies

ISP proxies delivered what we consider the best balance of speed and reliability this quarter. With a median TTFB of 385ms and a 96.1% success rate, they occupy a middle ground that works well for most use cases.

ISP proxy performance by session duration:

Session Length	Median TTFB	Success Rate
Single request	385ms	96.1%
1-5 min session	362ms	96.8%
5-30 min session	378ms	95.4%
30+ min session	410ms	93.2%

The slight degradation on very long sessions aligns with expectations: extended sessions from a single IP increase the chance of detection on protected sites. Rotating ISP proxies every 10-15 minutes appears to be the optimal strategy based on these numbers.

Results by Geography

Geographic location of both the proxy exit node and the target server significantly impacts performance. We measured response times from each of our five anchor servers to targets in four major regions.

Cross-Region Response Times (Residential Proxies, Median TTFB)

Proxy Exit Region	US Targets	EU Targets	Asia Targets	S. America Targets
United States	980ms	1,640ms	2,180ms	1,780ms
Europe	1,580ms	920ms	1,940ms	2,100ms
Asia-Pacific	2,240ms	2,020ms	860ms	2,680ms
South America	1,820ms	2,200ms	2,740ms	1,020ms

Key geographic findings:

Same-region pairings consistently outperformed cross-region by 40-60%.
US-to-EU routes showed the least cross-region penalty (roughly 1.7x slower vs. same-region), reflecting mature submarine cable infrastructure.
Asia-to-South-America was the worst-performing corridor at 2,740ms median, over 3x the same-region baseline.
Southeast Asian proxies showed higher variance than other Asia-Pacific locations, with P95 response times 2.2x the median compared to 1.8x for Japan/Korea exits.

Success Rates by Exit Region

Proxy Exit Region	Overall Success Rate	Best Category	Worst Category
United States	95.2%	News (98.1%)	Social media (91.4%)
Europe (DACH)	94.8%	Search engines (97.6%)	E-commerce (90.8%)
Europe (UK)	95.1%	News (97.8%)	Travel (90.2%)
Japan/Korea	93.6%	Search engines (96.4%)	Social media (89.1%)
Southeast Asia	92.8%	E-commerce (95.2%)	Travel (87.3%)
Brazil	93.1%	E-commerce (96.0%)	Social media (88.7%)

US and UK exit nodes showed the highest overall success rates, likely because the majority of tested target sites are English-language and US/UK-hosted, meaning the geo-match between proxy and target was strongest.

Results by Target Site Category

Detailed Performance Matrix (All Proxy Types, Median TTFB / Success Rate)

Target Category	Residential	Datacenter	Mobile	ISP
E-commerce	1,180ms / 95.8%	262ms / 85.7%	1,720ms / 98.1%	398ms / 96.4%
Social media	1,340ms / 92.1%	285ms / 82.3%	1,840ms / 96.8%	425ms / 94.2%
Search engines	1,020ms / 96.4%	218ms / 91.2%	1,540ms / 98.6%	355ms / 97.3%
News/media	980ms / 97.2%	195ms / 96.8%	1,480ms / 98.8%	328ms / 98.0%
Travel	1,280ms / 91.4%	310ms / 84.9%	1,920ms / 95.2%	442ms / 93.8%
Real estate	1,120ms / 94.8%	248ms / 91.4%	1,680ms / 97.0%	380ms / 95.6%

Notable Observations by Category

E-commerce sites have become the most polarized category. They are relatively easy to scrape with residential or mobile IPs (95-98% success) but increasingly difficult with datacenter IPs (85.7%). Amazon in particular has tightened its detection, with datacenter success rates dropping to 78.4% on product pages compared to 86.2% six months ago.

Social media platforms remain the hardest category overall. Even mobile proxies, which excel everywhere else, see their lowest success rates here (96.8%). Instagram and LinkedIn are the primary drivers of failures, with both platforms employing behavioral analysis that goes beyond simple IP reputation.

Search engines showed an interesting split. Google’s success rates have declined across all proxy types except mobile, while Bing and DuckDuckGo remain largely accessible. Google SERP scraping with datacenter proxies fell below 85% for the first time in our testing history.

News and media sites remain the easiest targets across the board, with even datacenter proxies maintaining a 96.8% success rate. Most news sites prioritize accessibility and SEO over bot prevention.

Travel sites (Booking.com, Expedia, airline sites) continue to be among the most aggressive with anti-bot measures. They showed the largest variance between proxy types, with a 13.9 percentage point gap between mobile (95.2%) and datacenter (84.9%) success rates.

Throughput Analysis

For tasks requiring large data transfers (downloading full pages, images, or API responses), throughput matters as much as latency.

Average Throughput by Proxy Type

Proxy Type	Median Throughput	P10 (Slow End)	P90 (Fast End)
Datacenter	18.6 Mbps	8.2 Mbps	42.0 Mbps
ISP	12.4 Mbps	4.8 Mbps	28.6 Mbps
Residential	4.2 Mbps	1.1 Mbps	12.8 Mbps
Mobile 4G	2.1 Mbps	0.6 Mbps	6.4 Mbps
Mobile 5G	5.8 Mbps	1.8 Mbps	16.2 Mbps

Datacenter proxies deliver roughly 4.4x the throughput of residential proxies at the median. This makes them significantly more efficient for bulk data collection from sites where they maintain high success rates. For bandwidth-heavy tasks on protected sites, ISP proxies offer a strong compromise at 12.4 Mbps median throughput with much better success rates than datacenter.

Key Findings and Surprises

1. The Datacenter Decline Is Accelerating

Datacenter proxy success rates dropped by an average of 3.4 percentage points across all target categories compared to Q3 2025. The decline is not uniform; news sites and APIs remain largely unaffected, while e-commerce and social media are driving most of the decline. For teams still relying heavily on datacenter proxies for e-commerce scraping, a migration plan is worth considering.

2. 5G Mobile Proxies Are a Game-Changer

The 38.5% speed improvement from 5G over 4G mobile proxies was the largest single performance jump we measured this quarter. At 1,180ms median TTFB, 5G mobile proxies are now faster than the average residential proxy (1,140ms at median) while maintaining the highest success rates in the industry. The main limitation remains availability: 5G exit nodes are concentrated in the US, South Korea, Japan, and parts of Western Europe.

3. ISP Proxies Are the New Default Recommendation

For the first time in our benchmarking history, we are positioning ISP proxies as the default recommendation for most use cases. At 385ms median TTFB and 96.1% success rate, they offer the best balance of speed, reliability, and cost-effectiveness. They are faster than residential proxies by 3x while maintaining comparable success rates.

4. Time-of-Day Effects Are Real but Modest

We observed a 6-12% performance improvement during off-peak hours (02:00-06:00 local time of the target server) across all proxy types. Residential proxies showed the largest time-of-day variance, likely because peer availability fluctuates with user activity patterns. The effect is real but rarely large enough to justify scheduling all scraping during off-peak windows.

5. Geographic Matching Matters More Than Proxy Type for Speed

Choosing a proxy exit node in the same region as your target server yields a larger speed improvement than switching proxy types. A residential proxy with perfect geo-matching (same country) was faster than an ISP proxy with poor geo-matching (different continent) in 72% of our test comparisons.

Recommendations Based on Data

By Use Case

Use Case	Recommended Proxy Type	Why
E-commerce scraping	Residential or ISP	Best success rate vs. cost balance
Social media monitoring	Mobile 4G/5G	Highest success on hardest targets
SERP tracking	ISP	Fast enough for real-time, high success
News aggregation	Datacenter	Speed matters, success rate is high
Travel price monitoring	Mobile or Residential	Travel sites block datacenter aggressively
Ad verification	Residential	Needs to mimic real user locations
API access	Datacenter	Speed is paramount, detection is minimal
General web scraping	ISP	Best all-around balance

Cost-Performance Optimization

Not every task needs the most expensive proxy type. Use our proxy cost calculator to estimate your spend based on these benchmarks, and check our proxy speed comparison tool for real-time performance data.

A hybrid approach works well for many teams:

Start with ISP proxies as your baseline for most targets.
Upgrade to mobile for high-value targets with aggressive detection.
Use datacenter for speed-sensitive, low-detection targets like APIs and news sites.
Deploy residential when you need broad geographic coverage or large IP diversity.

Methodology Transparency and Limitations

We believe in full transparency about the limitations of this benchmark:

Provider selection: We tested 8 residential, 6 datacenter, 5 mobile, and 5 ISP providers. These were selected based on market share and availability, but they do not represent the entire market. Results may differ with other providers.
Aggregated data: Performance figures represent aggregated medians across all tested providers within each category. Individual providers may significantly outperform or underperform these benchmarks.
Geographic bias: Our target site list skews toward English-language and Western sites. Performance on Chinese, Russian, or Arabic-language sites may differ significantly.
Temporal snapshot: These results reflect Q1 2026 conditions. Anti-bot systems and proxy networks evolve continuously. We recommend retesting quarterly.
No provider naming: We deliberately do not name or rank individual providers. This benchmark evaluates proxy types and categories, not specific vendors.

Data in this report is aggregated from independent testing conducted by DataResearchTools.com, publicly available provider benchmarks, and anonymized performance data contributed by users of our testing tools. Raw data files are available upon request for academic and research purposes.

About This Benchmark Series

This is the fourth installment of our quarterly proxy speed benchmark series. Previous reports are available in our archives. We conduct these benchmarks as a free resource for the proxy and web scraping community, with the goal of providing actionable, vendor-neutral performance data.

If you have suggestions for future benchmark categories or would like to contribute anonymized performance data from your own infrastructure, contact our research team at research@dataresearchtools.com.

Last updated: March 2026. Next benchmark scheduled for publication: July 2026 (Q2 2026 data).