proxy-benchmark
github.com/nodemaven/proxy-benchmark- Category
- Developer Tools
- Rank
- No. 1323Tools index
Previous survey · No. 1382 ·
- Pricing
- Open Source
- Type
- TOOL
- Use case
- Software Testing & Quality
- GitHub
- 69 stars
- Date
About
proxy-benchmark is a measurement harness, built and maintained by proxy vendor NodeMaven, that tests proxy providers, browser engines, and scraping targets to help identify whether a failed request is caused by the proxy, the browser, the host machine, or the target site. Every reported number is generated from committed run files in the repository so results can be independently recomputed rather than taken on faith.
At a glance
proxy-benchmark helps distinguish a proxy problem from a browser, host or target-site problem. It runs controlled combinations in one time window and writes a JSONL record for each attempt. You can use a direct connection first and bring your own proxy provider later.
Good fit
Useful when changing proxies or browsers has become guesswork. A direct control and interleaved comparisons make the question more specific: did the result change with the proxy path while other conditions stayed the same? The saved attempt parameters and response evidence also let you revisit how a result was classified.
Consider something else when
You want a permanent leaderboard of the best proxy or a guarantee that blocked requests will succeed. Host, target and time affect the findings. The maintainer, NodeMaven, sells proxies; read its published results as reproducible vendor research rather than an independent purchasing verdict.
Start here
From a checkout and Python virtual environment, install requirements-core.txt for the plain HTTP client. Use scripts/benchmark.py with --dry-run --engines http --targets bing_serp --queries 5 --batch 5 --direct --preset none to inspect the plan. A small direct run then establishes a baseline before adding a paid proxy or browser engine.
Cost and requirements
MIT software; the guide specifies Python 3.11 or newer. Core requirements do not install browsers. Browser engines add packages and roughly 2 GB of disk, while proxy traffic and any rented host are separate costs. Inspect the plan before spending metered traffic, and keep provider credentials out of committed files.
Limits to know
The published pass rate counts successful responses over judged attempts and excludes harness/path errors. Check error counts as well as that percentage. By default, failed response bodies and a sample of successful ones can be retained locally; --no-bodies saves storage but removes the ability to reclassify those responses later. Short or unmatched cells are weak comparison evidence.
Sources checked 2026-10-30. Source-based assessment.
- github.com/nodemaven/proxy-benchmark/blob/main/README.md
- github.com/nodemaven/proxy-benchmark/blob/main/pyproject.toml
- github.com/nodemaven/proxy-benchmark/blob/main/docs/quickstart.md
- github.com/nodemaven/proxy-benchmark/blob/main/requirements-core.txt
- github.com/nodemaven/proxy-benchmark/blob/main/nmbench/config.py
- github.com/nodemaven/proxy-benchmark/blob/main/scripts/benchmark.py
- github.com/nodemaven/proxy-benchmark/blob/main/RESULTS.md
- github.com/nodemaven/proxy-benchmark/blob/main/LICENSE
Why it made the leaderboard
Lets engineers evaluating scraping/proxy infrastructure isolate the actual cause of failed requests instead of guessing, with every benchmark number reproducible from committed raw run data.
Tags
Tech Stack
Media
Comments (0)
No comments yet
Editorially curated, with community endorsements as a secondary signal. Corrections welcome.