Hong Kong SAR China

Real home IPs

Hong Kong SAR China residential proxies

Hong Kong residential proxies are real IP addresses from Hong Kong households, assigned by consumer ISPs such as HKT, Hong Kong Broadband (HKBN), HGC and SmarTone. They let you browse, scrape and verify local storefronts, pricing and content as a genuine in-city user — across Central, Causeway Bay, Shatin and Tsuen Wan — instead of from a datacenter IP that HKTVmall, Fortress and local geo-gates flag and block on sight.

Try for $1.99/GB Sign in with Google

HTTP / SOCKS5City & ISP targetingTraffic never expiresVerified clean monthly

36,029IPs in pool

25+Cities covered

22ISP networks

150 msMedian latency

99.3%Strictly clean

Hong Kong SAR China residential proxies, at a glance

Hong Kong SAR China residential proxies are real IP addresses from Hong Kong households, assigned by consumer ISPs. They let you browse, scrape and verify Hong Kong websites, pricing and content as a genuine in-country user — instead of from a datacenter IP that local sites flag and block.

Pool type: Residential, multi-ISP
Cities covered: Central, Kwai Chung, Tsing Yi
Top ISPs: HKT · HKBN · HGC
Protocols: HTTP, SOCKS5
Rotation: Per-request or sticky
Geo-accuracy: 99.87%
Pool cleanliness: 99.3% clean
Pricing: from $2.00/GB
Last measured: 2026-06-23
Support: 24/7

Why teams pick our Hong Kong pool

Real residential IPs, precise targeting, and quality you don't have to take on faith.

Real residential IPs

Genuine home connections from Hong Kong consumer ISPs — not datacenter ranges that get blocked on sight.

City & ISP targeting

Target individual cities and carriers where pool depth allows — down to the provider.

Rotating or sticky

Per-request rotation for scale, or sticky sessions up to ~72h for multi-step flows.

HTTP & SOCKS5

Works with any client, scraper or automation stack out of the box.

Ethically sourced

IPs from an opt-in residential network — real users who consented to share bandwidth.

Verified clean, monthly

We re-measure the pool every month and publish the numbers, so they stay current.

Quality you can verify

Every month we run the same benchmark — a random sample of IPs scored against Scamalytics, latency tested from a neutral VPS, and geo-accuracy checked across independent databases. The figures below reflect production, not a best-case lab run.

Cleanliness Scamalytics · 100,000 IPs

Strictly clean99.3%

Datacenter contamination0%

On any blacklist0.19%

Low fraud-risk (0–19)99.2%

Geo-accuracy99.87%

Latency & speed VPS → Cloudflare

Median latency150ms

p90 / p95 latency194 / 245 ms

Connection success98.8%

Throughput (median / p95)2.59 / 7.22 MB/s

Large-file completion100%

Network composition

Connection type100% residential

ISP networks (ASN)22

Rotation modesper-request · sticky

Top carriersHKT · HKBN · HGC

Measured on 100,000 addresses from a neutral VPS against Cloudflare's edge, accurate as of the last benchmark. A repeat run may differ by a small margin, but the figures fully reflect the order of magnitude and are a sound basis for judging whether the pool fits your use case. Per-target scraping success isn't published — it depends on your fingerprint and setup, not just the IP.

benchmark.py — open pool benchmark

The open benchmark behind the numbers above. It takes a random sample of pool IPs and measures four things — cleanliness (Scamalytics fraud-risk, datacenter, blacklist and geo-accuracy), latency (Cloudflare edge and ip-api round-trip), rotation (unique IPs, ASNs and /24 subnets) and throughput (download speed and time-to-first-byte) — then reports each one. It is published in full so the methodology behind every figure stays open to inspection.

# -*- coding: utf-8 -*-

"""
JETPROXY — POOL BENCHMARK

End-to-end audit of a residential proxy pool. A single run measures cleanliness,
latency, rotation and throughput, then prints every metric as one list.

  Cleanliness   fraud-risk, datacenter, blacklist and geo-accuracy (Scamalytics)
  Latency       Cloudflare edge (responsiveness) and ip-api origin (round-trip)
  Rotation      unique IPs / ASNs / subnets, success rate, country match
  Throughput    median & p95 download speed, time-to-first-byte, drop rate

Every stage size is a parameter, so the same script runs a quick check or a deep
audit:

    python benchmark.py --country ID
    python benchmark.py --quality 200 --downloads 50 --latency 1000 --rotation 5000 --country ID
    python benchmark.py --proxies proxies.txt --country ID --json result.json

Proxies   paste at the prompt (one per line) or pass --proxies FILE.
          host:port:user:pass  ·  host:port  ·  user:pass@host:port
The Scamalytics key for the cleanliness stage is read from the environment or
../.env (SCAMALYTICS_USER + SCAMALYTICS_KEY); without it that stage is skipped.
"""
from __future__ import annotations

import argparse
import asyncio
import csv
import ipaddress
import json
import os
import re
import statistics
import sys
import time
from collections import Counter
from datetime import datetime

try:
    sys.stdout.reconfigure(encoding="utf-8")
    sys.stderr.reconfigure(encoding="utf-8")
except Exception:
    pass

try:
    import aiohttp
    from rich.console import Console
    from rich.table import Table
    from rich import box
except ImportError:
    print("\n[!] Need: pip install aiohttp rich\n")
    sys.exit(1)

console = Console()

# ── endpoints ────────────────────────────────────────────────────────────────

IP_ENDPOINT = ("http://ip-api.com/json/?fields="
               "status,message,country,countryCode,regionName,city,isp,as,org,query")
CF_TRACE = "https://speed.cloudflare.com/cdn-cgi/trace"   # tiny edge response (~350 B)

FILE_MB = 10
DOWN_URL = f"https://speed.cloudflare.com/__down?bytes={FILE_MB * 1_000_000}"
SCAM_HOST = "https://api11.scamalytics.com/v3"            # US node (api12 = EU)


SCORE_CONCURRENCY = 10
DOWN_CONCURRENCY = 3   # downloads share the local uplink — keep low so the exit node is the bottleneck



# ── proxy parsing ──────────────────────────────────────────────────────────────

def parse_proxy(raw: str):
    """host:port | host:port:user:pass | user:pass@host:port | scheme://..."""
    raw = raw.strip()
    if not raw or raw.startswith("#"):
        return None
    scheme = "http"
    if "://" in raw:
        scheme, raw = raw.split("://", 1)
    user = pwd = host = port = None
    if "," in raw:
        parts = [p.strip() for p in raw.split(",")]
        if len(parts) == 4:
            host, port, user, pwd = parts
        elif len(parts) == 2:
            host, port = parts
        else:
            return None
    elif "@" in raw:
        cred, hostport = raw.rsplit("@", 1)
        if ":" in cred:
            user, pwd = cred.split(":", 1)
        host, port = hostport.rsplit(":", 1)
    else:
        parts = raw.split(":")
        if len(parts) == 2:
            host, port = parts
        elif len(parts) == 4:
            host, port, user, pwd = parts
        else:
            return None
    try:
        port = int(port)
    except (ValueError, TypeError):
        return None
    return {"scheme": scheme, "host": host, "port": port, "user": user, "pwd": pwd,
            "url": f"{scheme}://{host}:{port}", "label": f"{host}:{port}"}


def subnet24(ip: str) -> str:
    try:
        return str(ipaddress.ip_network(ip + "/24", strict=False))
    except ValueError:
        return ""


def pct(values, p):
    if not values:
        return 0.0
    s = sorted(values)
    k = min(len(s) - 1, max(0, int(round(p / 100 * (len(s) - 1)))))
    return s[k]


# ── ip-api request (identity + geo + round-trip) ───────────────────────────────

def _fail(idx, proxy, err, latency):
    return {"idx": idx, "ok": False, "proxy": proxy["label"], "error": err, "latency": latency}


async def _one_attempt(session, proxy, idx, timeout):
    t0 = time.perf_counter()
    auth = aiohttp.BasicAuth(proxy["user"], proxy["pwd"]) if proxy["user"] else None
    try:
        async with session.get(IP_ENDPOINT, proxy=proxy["url"], proxy_auth=auth,
                               timeout=aiohttp.ClientTimeout(total=timeout)) as r:
            latency = (time.perf_counter() - t0) * 1000
            if r.status != 200:
                return _fail(idx, proxy, f"HTTP {r.status}", latency)
            data = await r.json(content_type=None)
            if data.get("status") != "success":
                return _fail(idx, proxy, data.get("message", "api error"), latency)
            ip = data.get("query", "")
            asn = data.get("as", "").split(" ")[0] if data.get("as") else ""
            return {"idx": idx, "ok": True, "proxy": proxy["label"], "ip": ip,
                    "country": data.get("country", ""), "cc": data.get("countryCode", ""),
                    "region": data.get("regionName", ""), "city": data.get("city", ""),
                    "isp": data.get("isp") or data.get("org", ""), "asn": asn,
                    "org": data.get("org") or data.get("as", ""),
                    "subnet": subnet24(ip), "latency": latency, "error": ""}
    except asyncio.TimeoutError:
        return _fail(idx, proxy, "timeout", (time.perf_counter() - t0) * 1000)
    except Exception as e:
        return _fail(idx, proxy, type(e).__name__, (time.perf_counter() - t0) * 1000)


async def fetch(session, proxy, sem, idx, timeout, max_retries=2):
    async with sem:
        res = None
        for attempt in range(1, max_retries + 2):
            res = await _one_attempt(session, proxy, idx, timeout)
            if res["ok"]:
                res["attempts"] = attempt
                res["first_try_ok"] = (attempt == 1)
                return res
        res["attempts"] = max_retries + 1
        res["first_try_ok"] = False
        return res


# ── Cloudflare edge probe (responsiveness + nearest datacenter) ────────────────

async def cf_probe(proxy, sem, timeout):
    """One warm-up request opens the tunnel + TLS, then the next request on the same
    kept-alive connection is timed — so the figure is the edge round-trip, not the
    one-off connection setup."""
    async with sem:
        auth = aiohttp.BasicAuth(proxy["user"], proxy["pwd"]) if proxy["user"] else None
        try:
            async with aiohttp.ClientSession(connector=aiohttp.TCPConnector(limit=2, ssl=False)) as sess:
                async with sess.get(CF_TRACE, proxy=proxy["url"], proxy_auth=auth,
                                    timeout=aiohttp.ClientTimeout(total=15)) as w:
                    await w.read()
                t0 = time.perf_counter()
                async with sess.get(CF_TRACE, proxy=proxy["url"], proxy_auth=auth,
                                    timeout=aiohttp.ClientTimeout(total=timeout)) as r:
                    txt = await r.text()
                    rtt = (time.perf_counter() - t0) * 1000
                    if r.status != 200:
                        return {"ok": False, "error": f"HTTP {r.status}"}
                    d = {}
                    for ln in txt.splitlines():
                        if "=" in ln:
                            k, v = ln.split("=", 1)
                            d[k] = v
                    return {"ok": True, "rtt": rtt, "ip": d.get("ip", ""),
                            "loc": d.get("loc", ""), "colo": d.get("colo", "")}
        except asyncio.TimeoutError:
            return {"ok": False, "error": "timeout"}
        except Exception as e:
            return {"ok": False, "error": type(e).__name__}


# ── download (throughput + TTFB + mid-transfer drops) ──────────────────────────

async def download(proxy, sem, idx, timeout, retries):
    async with sem:
        auth = aiohttp.BasicAuth(proxy["user"], proxy["pwd"]) if proxy["user"] else None
        last_err = ""
        dropped_at = []
        for attempt in range(1, retries + 2):
            try:
                async with aiohttp.ClientSession(connector=aiohttp.TCPConnector(limit=4, ssl=False)) as sess:
                    country = "?"
                    try:
                        async with sess.get(CF_TRACE, proxy=proxy["url"], proxy_auth=auth,
                                            timeout=aiohttp.ClientTimeout(total=15)) as tr:
                            for ln in (await tr.text()).splitlines():
                                if ln.startswith("loc="):
                                    country = ln[4:].strip() or "?"
                    except Exception:
                        pass
                    t0 = time.perf_counter()
                    async with sess.get(DOWN_URL, proxy=proxy["url"], proxy_auth=auth,
                                        timeout=aiohttp.ClientTimeout(total=timeout)) as r:
                        if r.status != 200:
                            last_err = f"HTTP {r.status}"
                            continue
                        first = None
                        nbytes = 0
                        try:
                            async for chunk in r.content.iter_chunked(65536):
                                if first is None:
                                    first = time.perf_counter()
                                nbytes += len(chunk)
                        except Exception as stream_err:
                            if nbytes > 0:
                                dropped_at.append(nbytes)
                                last_err = f"dropped@{nbytes // 1_000_000}MB"
                            else:
                                last_err = type(stream_err).__name__
                            continue
                        end = time.perf_counter()
                        if nbytes == 0:
                            last_err = "empty"
                            continue
                        ttfb = (first - t0) * 1000 if first else 0
                        dl_time = (end - first) if first and end > first else (end - t0)
                        mbps = (nbytes / 1_000_000) / dl_time if dl_time > 0 else 0
                        return {"idx": idx, "ok": True, "country": country, "bytes": nbytes,
                                "ttfb_ms": ttfb, "mbps": mbps, "dropped": bool(dropped_at), "error": ""}
            except asyncio.TimeoutError:
                last_err = "timeout"
            except Exception as e:
                last_err = type(e).__name__
        return {"idx": idx, "ok": False, "country": "?", "bytes": 0, "ttfb_ms": 0,
                "mbps": 0, "dropped": bool(dropped_at), "error": last_err}


# ── Scamalytics scoring ────────────────────────────────────────────────────────

def load_creds():
    user = os.environ.get("SCAMALYTICS_USER")
    key = os.environ.get("SCAMALYTICS_KEY")
    if user and key:
        return user.strip(), key.strip()
    found = {}
    for path in ("../.env", "../.env.local", ".env"):
        p = os.path.join(os.path.dirname(__file__), path)
        if os.path.exists(p):
            with open(p, encoding="utf-8") as f:
                for line in f:
                    m = re.match(r'\s*(SCAMALYTICS_USER|SCAMALYTICS_KEY)\s*=\s*"?([^"\s]+)"?', line)
                    if m:
                        found[m.group(1)] = m.group(2)
    return found.get("SCAMALYTICS_USER"), found.get("SCAMALYTICS_KEY")


def parse_scam(ip, data):
    s = data.get("scamalytics", {})
    if s.get("status") != "ok":
        return {"ip": ip, "ok": False, "error": s.get("status") or "bad_response"}
    ext = data.get("external_datasources", {})
    score = s.get("scamalytics_score")
    risk = s.get("scamalytics_risk", "?")
    px = s.get("scamalytics_proxy", {})
    is_dc = bool(px.get("is_datacenter"))
    is_vpn = bool(px.get("is_vpn"))
    fh = bool(ext.get("firehol", {}).get("is_proxy"))
    ips_bl = bool(ext.get("ipsum", {}).get("ip_blacklisted"))
    sh = bool(ext.get("spamhaus_drop", {}).get("ip_blacklisted"))
    x4 = bool(ext.get("x4bnet", {}).get("is_blacklisted_spambot"))
    on_bl = bool(s.get("is_blacklisted_external")) or fh or ips_bl or sh or x4
    cc = (ext.get("dbip", {}).get("ip_country_code")
          or ext.get("ipinfo", {}).get("ip_country_code")
          or ext.get("maxmind_geolite2", {}).get("ip_country_code")
          or ext.get("ip2proxy_lite", {}).get("ip_country_code") or "?")
    isp = (ext.get("dbip", {}).get("isp_name")
           or ext.get("ip2proxy_lite", {}).get("isp_name")
           or s.get("scamalytics_isp") or "?")
    asn = ext.get("ipinfo", {}).get("asn") or ext.get("ip2proxy_lite", {}).get("asn") or "?"
    return {"ip": ip, "ok": True, "error": "", "score": score if isinstance(score, int) else 0,
            "risk": risk, "is_dc": is_dc, "is_vpn": is_vpn, "on_bl": on_bl,
            "country": cc, "isp": isp, "asn": asn,
            "credits": s.get("credits", {}).get("remaining"),
            "strict_clean": (risk == "low" and not is_dc and not on_bl)}


async def score_one(session, user, key, ip, sem):
    async with sem:
        try:
            async with session.get(f"{SCAM_HOST}/{user}/", params={"key": key, "ip": ip},
                                   timeout=aiohttp.ClientTimeout(total=20)) as r:
                return parse_scam(ip, await r.json(content_type=None))
        except Exception as e:
            return {"ip": ip, "ok": False, "error": type(e).__name__}


# ── stage runners ──────────────────────────────────────────────────────────────

def progress(tag, done, total):
    step = max(1, total // 10)
    if done % step == 0 or done == total:
        console.print(f"   [{tag}] {done}/{total}", highlight=False)


async def run_ipapi(proxies, n, concurrency, timeout, retries, tag):
    sem = asyncio.Semaphore(concurrency)
    conn = aiohttp.TCPConnector(limit=concurrency * 3, ssl=False)
    out = []
    async with aiohttp.ClientSession(connector=conn) as http:
        tasks = [asyncio.create_task(fetch(http, proxies[i % len(proxies)], sem, i, timeout, retries))
                 for i in range(n)]
        for done, coro in enumerate(asyncio.as_completed(tasks), 1):
            out.append(await coro)
            progress(tag, done, n)
    return out


async def run_cf(proxies, n, concurrency, timeout, tag):
    sem = asyncio.Semaphore(concurrency)
    tasks = [asyncio.create_task(cf_probe(proxies[i % len(proxies)], sem, timeout))
             for i in range(n)]
    out = []
    for done, coro in enumerate(asyncio.as_completed(tasks), 1):
        out.append(await coro)
        progress(tag, done, n)
    return out


async def run_download(proxies, n, timeout, retries, tag):
    sem = asyncio.Semaphore(DOWN_CONCURRENCY)
    tasks = [asyncio.create_task(download(proxies[i % len(proxies)], sem, i, timeout, retries))
             for i in range(n)]
    out = []
    for done, coro in enumerate(asyncio.as_completed(tasks), 1):
        out.append(await coro)
        progress(tag, done, n)
    return out


async def score_ips(ips, user, key, tag):
    sem = asyncio.Semaphore(SCORE_CONCURRENCY)
    out = []
    async with aiohttp.ClientSession() as session:
        tasks = [asyncio.create_task(score_one(session, user, key, ip, sem)) for ip in ips]
        for done, coro in enumerate(asyncio.as_completed(tasks), 1):
            out.append(await coro)
            progress(tag, done, len(ips))
    return out


# ── input ──────────────────────────────────────────────────────────────────────

def read_proxies(path):
    if path:
        with open(path, encoding="utf-8") as f:
            lines = f.readlines()
    else:
        console.print("Paste proxies (one per line), then an empty line:")
        lines = []
        while True:
            try:
                ln = input()
            except EOFError:
                break
            if ln.strip() == "":
                if lines:
                    break
                continue
            lines.append(ln)
    return [p for p in (parse_proxy(l) for l in lines) if p]


def fmt(v, suffix=""):
    return f"{v}{suffix}"


# ── main ─────────────────────────────────────────────────────────────────────

def build_report(M):
    t = Table(title="POOL BENCHMARK — RESULTS", box=box.DOUBLE_EDGE, header_style="bold cyan",
              title_style="bold green")
    t.add_column("Metric", style="cyan")
    t.add_column("Value", style="bold white", justify="right")
    for section, rows in M["sections"]:
        t.add_row(f"[bold yellow]{section}[/]", "")
        for k, v in rows:
            t.add_row(f"  {k}", str(v))
    console.print()
    console.print(t)


async def run(args, proxies):
    M = {"label": args.label, "country": args.country or None,
         "timestamp": datetime.now().isoformat(timespec="seconds"),
         "proxies_input": len(proxies), "stages": {}}
    user = key = None

    # 1 — cleanliness (Scamalytics)

    clean_rows = []
    if not args.no_quality and args.quality > 0:
        user, key = load_creds()
        if not user or not key:
            console.print("[yellow]Scamalytics key not found — skipping cleanliness stage.[/]")
        else:
            console.rule(f"[bold]Cleanliness — {args.quality} IPs[/]")
            collected = await run_ipapi(proxies, int(args.quality * 1.6) + 10,
                                        args.concurrency, 15, 1, "collect")
            uniq = {}
            for r in collected:
                if r.get("ok") and r.get("ip") and r["ip"] not in uniq:
                    uniq[r["ip"]] = r
            ips = list(uniq.keys())[:args.quality]
            scored = await score_ips(ips, user, key, "score")
            ok = [r for r in scored if r.get("ok")]
            n = len(ok)
            if n:
                strict = sum(r["strict_clean"] for r in ok) / n * 100
                low = sum(1 for r in ok if r["risk"] == "low") / n * 100
                dc = sum(r["is_dc"] for r in ok) / n * 100
                bl = sum(r["on_bl"] for r in ok) / n * 100
                vpn = sum(r["is_vpn"] for r in ok)
                geo = None
                if args.country:
                    geo = sum(1 for r in ok if (r["country"] or "").upper() == args.country.upper()) / n * 100
                M["stages"]["cleanliness"] = {
                    "n": n, "strict_clean_pct": round(strict, 1), "low_risk_pct": round(low, 1),
                    "datacenter_pct": round(dc, 1), "blacklist_pct": round(bl, 1),
                    "geo_accuracy_pct": round(geo, 1) if geo is not None else None, "vpn": vpn}
                clean_rows = [("Strictly clean", f"{strict:.1f}%"),
                              ("Low fraud-risk (0–19)", f"{low:.1f}%"),
                              ("Datacenter", f"{dc:.1f}%"),
                              ("Blacklisted", f"{bl:.1f}%")]
                if geo is not None:
                    clean_rows.append((f"Geo-accuracy ({args.country.upper()})", f"{geo:.1f}%"))
                # per-IP CSV

                ts = datetime.now().strftime("%Y%m%d_%H%M%S")
                with open(f"benchmark_ips_{ts}.csv", "w", newline="", encoding="utf-8") as f:
                    w = csv.writer(f)
                    w.writerow(["ip", "score", "risk", "is_datacenter", "is_vpn", "on_blacklist",
                                "country", "isp", "asn", "strict_clean"])
                    for r in scored:
                        if r.get("ok"):
                            w.writerow([r["ip"], r["score"], r["risk"], r["is_dc"], r["is_vpn"],
                                        r["on_bl"], r["country"], r["isp"], r["asn"], r["strict_clean"]])

    # 2 — throughput

    console.rule(f"[bold]Throughput — {args.downloads} downloads[/]")
    dl = await run_download(proxies, args.downloads, 60, 1, "download")
    dok = [r for r in dl if r["ok"]]
    dl_rows = []
    if dok:
        mbps = [r["mbps"] for r in dok]
        ttfb = [r["ttfb_ms"] for r in dok]
        nodrop = sum(1 for r in dl if r["ok"] and not r["dropped"]) / len(dl) * 100
        M["stages"]["throughput"] = {
            "n": len(dl), "ok": len(dok),
            "median_mbps": round(statistics.median(mbps), 2), "p95_mbps": round(pct(mbps, 95), 2),
            "median_ttfb_ms": round(statistics.median(ttfb)), "completed_no_drop_pct": round(nodrop, 1)}
        dl_rows = [("Median download", f"{statistics.median(mbps):.1f} MB/s ({statistics.median(mbps)*8:.0f} Mbps)"),
                   ("p95 download", f"{pct(mbps,95):.1f} MB/s ({pct(mbps,95)*8:.0f} Mbps)"),
                   ("Median TTFB", f"{statistics.median(ttfb):.0f} ms"),
                   ("Completed without drop", f"{nodrop:.0f}%")]

    # 3 — latency (Cloudflare edge + ip-api origin)

    console.rule(f"[bold]Latency — {args.latency} probes[/]")
    cf = await run_cf(proxies, args.latency, args.concurrency, 15, "cf")
    cfok = [r for r in cf if r["ok"]]
    lat_rows = []
    if cfok:
        rtts = [r["rtt"] for r in cfok]
        colo = Counter(r["colo"] for r in cfok if r["colo"]).most_common(1)
        M["stages"].setdefault("latency", {})
        M["stages"]["latency"]["cloudflare"] = {
            "n": len(cfok), "p50_ms": round(statistics.median(rtts)),
            "p90_ms": round(pct(rtts, 90)), "p95_ms": round(pct(rtts, 95)),
            "top_colo": colo[0][0] if colo else None}
        lat_rows += [("Cloudflare edge — p50", f"{statistics.median(rtts):.0f} ms"),
                     ("Cloudflare edge — p90", f"{pct(rtts,90):.0f} ms"),
                     ("Cloudflare edge — p95", f"{pct(rtts,95):.0f} ms")]
        if colo:
            lat_rows.append(("Nearest datacenter", colo[0][0]))

    # 4 — rotation (also yields ip-api round-trip latency)

    console.rule(f"[bold]Rotation — {args.rotation} requests[/]")
    rot = await run_ipapi(proxies, args.rotation, args.concurrency, args.timeout, 2, "rotation")
    rok = [r for r in rot if r["ok"]]
    rot_rows = []
    if rok:
        uips = {r["ip"] for r in rok if r.get("ip")}
        uasn = {r["asn"] for r in rok if r.get("asn")}
        usub = {r["subnet"] for r in rok if r.get("subnet")}
        success = len(rok) / len(rot) * 100
        first = sum(1 for r in rok if r.get("first_try_ok")) / len(rot) * 100
        cmatch = None
        if args.country:
            cmatch = sum(1 for r in rok if (r["cc"] or "").upper() == args.country.upper()) / len(rok) * 100
        ipapi_rtts = [r["latency"] for r in rok]
        M["stages"]["rotation"] = {
            "n": len(rot), "success_pct": round(success, 1), "first_try_pct": round(first, 1),
            "unique_ips": len(uips), "unique_asns": len(uasn), "unique_subnets": len(usub),
            "country_match_pct": round(cmatch, 2) if cmatch is not None else None}
        M["stages"].setdefault("latency", {})
        M["stages"]["latency"]["ip_api"] = {
            "n": len(ipapi_rtts), "p50_ms": round(statistics.median(ipapi_rtts)),
            "p90_ms": round(pct(ipapi_rtts, 90))}
        lat_rows += [("ip-api origin — p50", f"{statistics.median(ipapi_rtts):.0f} ms"),
                     ("ip-api origin — p90", f"{pct(ipapi_rtts,90):.0f} ms")]
        rot_rows = [("Success rate", f"{success:.1f}%"),
                    ("First-try success", f"{first:.1f}%"),
                    ("Unique exit IPs", f"{len(uips)}"),
                    ("Unique networks (ASN)", f"{len(uasn)}"),
                    ("Unique /24 subnets", f"{len(usub)}")]
        if cmatch is not None:
            rot_rows.append((f"Country match ({args.country.upper()})", f"{cmatch:.2f}%"))

    M["sections"] = [s for s in [
        ("CLEANLINESS", clean_rows), ("LATENCY", lat_rows),
        ("ROTATION", rot_rows), ("THROUGHPUT", dl_rows)] if s[1]]
    build_report(M)

    out_path = args.json or f"benchmark_{args.label}_{datetime.now().strftime('%Y%m%d_%H%M%S')}.json"
    M.pop("sections", None)
    with open(out_path, "w", encoding="utf-8") as f:
        json.dump(M, f, indent=2, ensure_ascii=False)
    console.print(f"\n[green]✓[/] Saved: [bold]{out_path}[/]")


def main():
    ap = argparse.ArgumentParser(description="Jetproxy pool benchmark")
    ap.add_argument("--quality", type=int, default=100, help="IPs scored for cleanliness")
    ap.add_argument("--downloads", type=int, default=50, help="download samples")
    ap.add_argument("--latency", type=int, default=1000, help="latency probes")
    ap.add_argument("--rotation", type=int, default=2000, help="rotation requests")
    ap.add_argument("--concurrency", type=int, default=20)
    ap.add_argument("--timeout", type=float, default=15.0)
    ap.add_argument("--country", default="", help="expected ISO code for geo-accuracy")
    ap.add_argument("--proxies", default="", help="proxy list file (else paste)")
    ap.add_argument("--label", default="pool")
    ap.add_argument("--json", default="")
    ap.add_argument("--no-quality", action="store_true", help="skip the cleanliness stage")
    args = ap.parse_args()

    console.print("\n[bold cyan]═══  JETPROXY — POOL BENCHMARK  ═══[/]\n")
    proxies = read_proxies(args.proxies)
    if not proxies:
        console.print("[red]No valid proxies.[/]")
        return
    console.print(f"[green]✓[/] Proxies loaded: [bold]{len(proxies)}[/]")
    asyncio.run(run(args, proxies))


if __name__ == "__main__":
    try:
        main()
    except KeyboardInterrupt:
        console.print("\n[yellow]Interrupted.[/]")

Target down to the city

«Avg. online» is the average number of distinct devices connected and available at the same moment — not a cumulative total. Grouped into High / Medium / Low.

RegionsCitiesProviders

Region	Latency	Avg. online	Pool
Kwai Tsing District	149 ms	~710	High
Central and Western	150 ms	~648	High
Kowloon City	149 ms	~203	High
Yuen Long District	151 ms	~182	High
Sha Tin	150 ms	~171	High
Eastern District	149 ms	~163	High
Kwun Tong District	151 ms	~139	High
Yau Tsim Mong District	153 ms	~129	Medium
Tsuen Wan District	151 ms	~128	Medium
North District	148 ms	~119	Medium
Sham Shui Po District	149 ms	~101	Medium
Wong Tai Sin District	150 ms	~94	Medium
Sai Kung District	153 ms	~89	Low
Tai Po District	151 ms	~76	Low
Kowloon	153 ms	~73	Low
Tuen Mun	151 ms	~65	Low
Islands	154 ms	~45	Low
Southern District	151 ms	~42	Low
Wan Chai	154 ms	~27	Low

City	Latency	Avg. online	Pool
Central	150 ms	~839	High
Kwai Chung	149 ms	~778	High
Tsing Yi	150 ms	~199	High
Hong Kong	153 ms	~136	High
Kwu Tung	150 ms	~126	High
Tin Shui Wai	151 ms	~93	High
Causeway Bay	148 ms	~79	High
Ho Man Tin	147 ms	~76	High
Tsuen Wan	153 ms	~69	High
Shatin	149 ms	~67	Medium
Tai Po	151 ms	~64	Medium
Sham Shui Po	149 ms	~64	Medium
Lam Tin	185 ms	~61	Medium
Tseung Kwan O	149 ms	~56	Medium
Hang Hau	155 ms	~56	Medium
Tai Kok Tsui	146 ms	~51	Medium
Mong Kok	161 ms	~49	Low
Kowloon Bay	149 ms	~45	Low
Yuen Long	146 ms	~45	Low
Chai Wan Kok	152 ms	~44	Low
Tung Chung	154 ms	~44	Low
Hung Hom	153 ms	~42	Low
Tuenmen	147 ms	~42	Low
Kam Tin	180 ms	~40	Low
Jordan	153 ms	~39	Low

Provider	Latency	Avg. online	Pool
HKTfixed	150 ms	~1,433	High
HKBNfixed	137 ms	~1,075	High
HGCfixed	150 ms	~349	High
CommuniLinkmobile	152 ms	~137	Medium
SmarTonemobile	201 ms	~110	Low
HKCfixed	153 ms	~96	Low
NOVAfixed	148 ms	~4	Low

Geo-accuracy — every IP sold as Hong Kong SAR China resolves to Hong Kong SAR China across three independent geo-databases (db-ip, MaxMind, ipinfo); high geo-accuracy is the signal local geo-gates rely on.

Hong Kong SAR China proxies by use case

Organised by what you're doing, not by platform. Local services are listed where they matter.

E-commerce & market research

Check storefronts, prices and local listings from inside the market.

Storefront monitoring
Local price tracking
Seller research

HKTVmallYOHO (友和)Fortress (豐澤)Broadway (百老滙)ParknShop (百佳)

Ad verification & performance

See how ads, creatives and SERP render to a local user — no manual switching.

Local SERP
Creative checks
Landing pages

WhatsAppFacebookInstagramWeChat

Travel & price intelligence

Capture flight and hotel fares the way a local customer sees them.

Geo-priced fares
Offer monitoring
Rate comparison

Trip.comKlookKKdayCathay PacificHK Express

Streaming & geo-verification

Verify region-locked catalogues and geo-gates from the right location.

Geo-checks
Catalogue verification
Regional availability

myTV SUPER (TVB)ViuNow E (PCCW)HBO GO AsiaNetflix

Classifieds & local listings

Monitor used-goods, autos and rentals across local classifieds.

Listing monitoring
Price tracking
Regional coverage

CarousellDCFever28car (易發車網)Geo Expat

Dev & automation

Residential IPs for pipelines, tests and permitted data collection via API.

Pipeline integration
API management
Sticky sessions

PythonScrapycURLGitHub

Beyond local sites, the same Hong Kong pool works for the major global platforms — Google, Meta, Amazon, YouTube, TikTok and Netflix — seen from the Hong Kong locale: local pricing, ads, catalogues, SERP and content.

Developer API

Generate Hong Kong SAR China-scoped endpoints in the dashboard, copy tested snippets, or automate proxy lists through the public API. City, ISP, sticky sessions and saved lists stay in one clean workflow.

API docs

1import requests
2 
3headers = {"x-api-key": "gpx_live_YOUR_KEY"}
4payload = {
5  "country_code": "HK",
6  "city": "Central",
7  "session_type": "sticky",
8  "quantity": 5,
9  "protocol": "socks5"
10}
11 
12response = requests.post(
13  "https://jetproxy.io/api/v1/generate",
14  headers=headers,
15  json=payload,
16  timeout=30
17)
18print(response.json()["proxies"])

Packages with no hidden terms

Higher volume means lower price per gigabyte. No subscriptions or hidden fees.

1 GB$3.70

$3.70 / GB

Select

5 GB$18.00

$3.60 / GB

Select

10 GB$34.50

$3.45 / GB

Select

Popular

25 GB$80.00

$3.20 / GB

Select

50 GB$147.50

$2.95 / GB

Select

100 GB$270.00

$2.70 / GB

Select

250 GB$612.50

$2.45 / GB

Select

500 GB$1,100.00

$2.20 / GB

Select

1 TB$2,000.00

$2.00 / GB

Select

Package

Price per GB

Total

1 GB

$3.70 / GB

$3.70Select

5 GB

$3.60 / GB

$18.00Select

10 GB

$3.45 / GB

$34.50Select

Popular

25 GB

$3.20 / GB

$80.00Select

50 GB

$2.95 / GB

$147.50Select

100 GB

$2.70 / GB

$270.00Select

250 GB

$2.45 / GB

$612.50Select

500 GB

$2.20 / GB

$1,100.00Select

1 TB

$2.00 / GB

$2,000.00Select

Available payment methods

Questions buyers actually ask

Still weighing whether the Hong Kong pool fits your use case? Our team answers real questions on Telegram, around the clock — before and after you buy.

Chat on Telegram

Will these work for scraping HKTVmall and Fortress?

That's the primary use case. The pool is genuine residential — the property local e-commerce anti-bot systems look for — so these IPs pass where datacenter ranges get blocked. Outcomes still depend on your fingerprint, headers and request pacing, so pair the proxy with a clean setup and a matching Hong Kong timezone (HKT, UTC+8).

Can I target a specific city like Central or Causeway Bay?

Yes, where pool depth allows — choose the city when you generate the proxy list. Our coverage tables show live concurrency for Central, Kwai Chung, Tsing Yi, Causeway Bay, Shatin and Tsuen Wan and more, so you can see which cities carry the deepest rotation before you commit.

Do the IPs geolocate inside Hong Kong?

Yes — the pool resolves to Hong Kong with 99.87% geo-accuracy against Cloudflare and across independent geo-databases in our last benchmark, with exit nodes confirmed at the Hong Kong (HKG) routing colo. High geo-accuracy is the underlying signal Hong Kong geo-gates rely on.

Can I hold one Hong Kong IP for a multi-step flow?

Yes. Use a sticky session to keep the same Hong Kong exit IP — average hold is around 5 hours and up to 72 hours in our 3-day measurement. For scale, rotate per request: the exit IP changes in about 98% of requests, usually down to a different provider, not just a neighbouring address.

Which Hong Kong carriers and cities are covered?

The pool spans 9 Hong Kong ISPs led by HKT, Hong Kong Broadband (HKBN), HGC, SmarTone and CommuniLink, across 19 regions and 25 cities — from Hong Kong Island (Central, Causeway Bay) to Kowloon (Sham Shui Po, Ho Man Tin) and the New Territories (Shatin, Tsuen Wan, Tai Po, Kwai Chung).