Oxylabs pricing vs Thordata: what proxies really cost a research project

If you searched for Oxylabs pricing, you almost certainly hit the same wall everyone else does: the marketing pages say starts from $6/GB and then ask you to talk to sales. This page does the arithmetic Oxylabs does not put plainly on its own site, prices the same three research-project volumes against Thordata, and — before either — tells you the case in which you should not be buying proxy bandwidth at all.

Every figure below is quoted from each vendor’s own public pricing pages, checked on 26 August 2026. Proxy list prices move often; re-check before you build a budget line on them.

Tip: try code CASRAI at checkout for 15% off, if the offer is currently active for this program — codes vary by vendor and aren’t guaranteed.

First, rule out the route that costs nothing

This is the part a page monetised by proxy referrals is not supposed to lead with, so we will lead with it. For most academic projects, neither Oxylabs nor Thordata is the right answer, because the data you want is already available through an official interface.

• Scholarly metadata and citations — Crossref’s REST API and OpenAlex both expose the full corpus openly, with no rate ceiling that a normal project will trouble and no terms-of-service exposure.
• Biomedical literature — NCBI E-utilities covers PubMed and PMC, free, with a documented rate limit and a registered API key that raises it.
• Government and statistical data — almost every national statistics office, funder and regulator now publishes a bulk download or an API precisely so you do not have to scrape the HTML.
• Platform data behind a wall — many providers run a formal research-access programme, and an institutional data-sharing agreement negotiated by your library or research office gets you cleaner data, on defensible terms, than any proxy pool will.

An official route wins on three axes that matter more to a research project than price: ethics (you are not circumventing an access control), reproducibility (a versioned API snapshot can be re-run by a reviewer; a scrape through a rotating residential pool frequently cannot — see reproducibility), and durability (an API does not break the week a site changes its markup). If you have not genuinely ruled these out, close this page and go do that first. Our guide to data collection methods is a better starting point than any vendor comparison.

Proxies earn their place in a narrower set of cases: the source publishes no API, the API omits the fields your design needs, the data is geographically partitioned and your research question is about that partitioning, or you are collecting at a scale that a single institutional IP cannot sustain without being rate-limited into uselessness. If that is you, the rest of this page is for you.

What Oxylabs actually costs per GB

Oxylabs publishes four residential-proxy plan tiers. As of 26 August 2026 they are:

Two things about that table matter more than the numbers in it.

The tiers are commitments, not meters. You buy the bucket, not the bytes. A project that consumes 50 GB in a month sits awkwardly between Basic and Advanced: buy Advanced and you have paid $500 for 50 GB of actual use, an effective $10/GB; buy Basic and top up the missing 30 GB at the Basic rate and you are nearer $250. Which of those you land on depends on top-up terms Oxylabs does not publish in full.

Above 1 TB there is no published rate at all. Corporate is the largest plan on the page; anything beyond it is a sales conversation. The honest reading is that a 5 TB buyer almost certainly negotiates below the $2.50/GB list rate — enterprise proxy pricing is discounted at volume as a matter of routine. We cannot tell you what that negotiated number is, and neither can anyone else who has not been quoted. That undisclosed number is the entire reason this search query exists.

Oxylabs also lists a Web Scraper API from $49/month, ISP proxies and mobile proxies as separate lines, and a free trial of the residential pool that is available once per customer and arranged through sales rather than self-service.

Thordata priced against the same volumes

Thordata prices residential traffic on a published volume slider rather than fixed plan buckets. As of 26 August 2026 the band runs from $2.00/GB at 1 GB down to $0.65/GB at 5,000 GB, with a headline “from $0.65/GB” against a $1.05/GB list. The minimum purchase is 1 GB. Datacenter and static ISP proxies are priced per IP from $0.75; mobile traffic runs $5.00/GB at 1 GB down to $2.20/GB at 500 GB. A SERP API is offered from roughly $0.70–$0.80 per 1,000 requests and a Web Scraper API from $0.50 per 1,000 results, each with a 5,000-unit free trial.

We are deliberately not inventing the intermediate rates. Thordata’s slider sets a specific price at 50 GB and at 500 GB; we are quoting only the two published endpoints and treating everything between them as bounded by those endpoints. Check the live slider for the exact figure at your volume before you budget.

See Thordata per-GB pricing →

The arithmetic at 50 GB, 500 GB and 5 TB

Three volumes that correspond to real research shapes: a pilot or a single-site study, a full year of a mid-sized collection, and a large longitudinal or multi-country panel.

Read the bottom row carefully, because it is where honest comparison stops. At 50 GB and 500 GB the published rates are directly comparable and Thordata is straightforwardly cheaper by a multiple. At 5 TB we are comparing a published price against a list rate nobody actually pays, and we will not pretend that is a like-for-like result. If you are buying at that scale, get the Oxylabs quote before you decide — the gap may narrow considerably.

Where Oxylabs earns its premium

A 3–5× price difference is not a mystery to be explained away; it buys real things, and for some projects those things are the deciding factor.

• Scraper and SERP infrastructure that is a product, not a line item. Oxylabs’ Web Scraper API and Web Unblocker are mature, documented, and maintained against the anti-bot systems they target. If your project’s cost centre is graduate-student hours spent fixing a brittle collector, the cheaper bandwidth can be a false economy.
• Compliance paperwork that survives contact with a legal office. Oxylabs publishes a formal ethical-sourcing position for its residential pool, maintains the documentation an institutional review will ask for, and has the corporate footprint to answer a due-diligence questionnaire. This is genuinely harder for smaller vendors to match.
• Contractual guarantees. Named account management, an SLA with numbers in it, and an escalation path that is a person rather than a chat widget.
• Pool scale. 175M+ IPs across 195+ countries, against Thordata’s claimed 100M across 190+.

None of that is decorative if your data collection sits behind a funder-committed deliverable with a date attached.

Start a costed Thordata pilot →

Pool size versus what a real scraping project consumes

Vendor pool sizes — 175M versus 100M IPs — are the headline number in every proxy comparison, and for most research projects they are close to irrelevant. Pool size governs how long you can rotate before an address repeats and how granularly you can target a city or an ASN. A study collecting from a few hundred domains, at polite rates, from a handful of countries, will never exhaust either pool.

What actually governs your bill is bytes, and the single biggest lever on bytes is whether you are rendering pages. A raw HTML fetch of a typical article page is tens of kilobytes. The same page through a headless browser, pulling images, fonts, analytics and ad scripts, is routinely 1–3 MB — a 30–50× multiplier applied to every request, billed as residential bandwidth. Blocking non-essential resource types in the browser context, or dropping to plain HTTP requests wherever the markup is server-rendered, will cut a proxy bill further than any vendor switch. Our comparison of Playwright vs Selenium covers where that control lives in each stack, and our guide to web scraping proxies covers the proxy-type choice itself.

The practical implication for this comparison: measure your real per-page byte cost on a small sample before you pick a tier at all. Teams routinely over-buy by an order of magnitude because they estimated from page count rather than payload.

Institutional procurement: invoicing, DPAs and whose terms your legal office will sign

This is where the cheaper vendor most often loses, and it is the axis a purely technical comparison misses entirely.

• Can they invoice? Many grant accounts cannot be spent on a personal card against a self-service checkout. If a vendor is card-only, the purchase may be administratively impossible regardless of price.
• Will they sign your data processing agreement? If your collection touches personal data — and web data more often does than researchers expect — your institution will require a DPA on its paper, or at minimum a vendor DPA its data protection officer will accept. A vendor that offers only click-through terms is a hard stop at many universities.
• Whose terms govern? Click-through terms of service with a foreign governing-law clause and an arbitration provision are frequently unsignable by a public institution. A negotiable master services agreement is a real, chargeable service.
• Security review. Expect a questionnaire, possibly SOC 2 or ISO 27001 evidence, and a sub-processor list.
• Renewal and audit. Multi-year grant budgets need a price that holds and an invoice trail that survives an audit.

Budget four to twelve weeks for this if the answer is not already “we have an existing agreement.” It is worth involving your research office before you pick a vendor, not after — a rejected DPA at month three of a twelve-month collection is a much more expensive problem than a higher per-GB rate.

Ethics and terms of service before you collect anything

Cheaper bandwidth does not lower the standard your work is held to. Before any collection run:

• Read robots.txt and honour it. It is not legally binding everywhere, but ignoring it is an editorial and ethical position you will have to defend to a reviewer, and increasingly to a journal’s research-integrity process.
• Rate-limit conservatively. Identify your crawler in the user-agent with a contact address, back off on errors, and stay well below any published limit. Degrading a service you are studying is a harm you caused.
• Get the review done first. Public web data is not automatically exempt from human-subjects oversight. If your collection touches identifiable individuals, aggregates into re-identifiable profiles, or covers a vulnerable population, that is an IRB or REC question — and a consent question — before it is a technical one. Ask your ethics office rather than deciding for yourself that it is exempt.
• Terms of service are a real constraint. Many sites prohibit automated collection outright. That does not always translate to legal liability, but it does translate to institutional risk, and your legal office would rather hear about it now.
• Write it into the plan. Collection method, lawful basis, retention period and deletion schedule belong in your data management plan, and reviewers increasingly look for them.

No proxy vendor absorbs any of this responsibility on your behalf, whatever their marketing implies about “compliant” scraping.

Run a costed pilot before you commit a grant budget line

The reliable way to avoid both over-buying and a mid-project vendor switch:

1. Sample 200–500 target pages representative of the full set — not the easy ones.
2. Measure bytes per successful record, not per request. Failures, retries and redirects all bill.
3. Record the success rate per vendor on the same sample, in the same week. A pool with a 60% success rate at $0.65/GB is not cheaper than one with 95% at $2.00/GB once retries are counted — this is the single most common costing error.
4. Multiply out to your full corpus, then add 40% for markup changes, blocks and re-runs across the collection period.
5. Price that number at both vendors, and get the Oxylabs quote if you land above 1 TB.
6. Take the total to your research office with the procurement questions above already asked.

Both vendors offer free trial traffic, which makes step 3 cost nothing but a day. Our Thordata review covers what that trial does and does not include.

Try Thordata free before committing grant funds →

Who should not buy Thordata

Stated plainly, because a comparison that never recommends the other vendor is not a comparison:

Choose Oxylabs if your project needs a signed enterprise agreement, a named support contact, a DPA your institution’s legal office will actually accept, or a guaranteed SLA standing behind a funder-committed deliverable. Its compliance and procurement apparatus is genuinely stronger, and at that point the premium is buying risk reduction, not bandwidth. The same applies if you are relying on a managed scraping API rather than raw proxies — Oxylabs’ is the more mature product.

Choose Thordata if you are price-sensitive, technically self-sufficient, buying at pilot-to-mid scale, and able to purchase without a negotiated institutional contract. On published rates it is materially cheaper at every volume where both vendors publish a number.

Choose neither if an official API, a bulk download or a negotiated data-sharing agreement can answer your research question. That remains, for most academic projects, the correct answer.

Frequently asked questions

Why does Oxylabs not publish a price for large volumes?

Its published plan tiers stop at Corporate ($2,500 for 1 TB). Anything above that is quoted by sales, which is standard for enterprise proxy vendors and lets the rate be negotiated per customer. It also means the published $2.50/GB is a ceiling for large buyers, not the real price — get a quote rather than budgeting from the list rate.

Is Thordata’s $0.65/GB the price I will actually pay?

Only at 5,000 GB. It is the bottom of a published band that starts at $2.00/GB for 1 GB, and it is quoted against a $1.05/GB list as a promotional rate. Check the live slider at your actual volume, and note that promotional pricing is not a durable basis for a multi-year grant budget.

Does a bigger proxy pool mean better data?

Rarely, for research workloads. Pool size affects rotation depth and geographic granularity. Your success rate and your bytes-per-record are what determine both data quality and cost, and both are dominated by your own collector design — particularly whether you render pages in a headless browser.

Can I put proxy costs on a grant?

Usually yes as a direct cost where the collection is integral to the funded work, subject to your funder’s rules and your institution’s procurement policy. The practical obstacles are administrative rather than allowability — invoicing, contract signature and data protection review. Confirm with your research office before the award budget is finalised.

Do I need ethics approval to scrape public web pages?

Sometimes, and it is not your call to make unilaterally. Public availability does not by itself make data exempt from human-subjects oversight, particularly where records are identifiable, re-identifiable in aggregate, or concern a vulnerable population. Ask your IRB or research ethics committee before collection begins.

Related CASRAI guidance

• Web scraping proxies for research data collection — proxy types, when each is appropriate, and the compliance framing.
• Thordata review: proxies for academic data collection — a fuller look at the vendor in this comparison.
• OpenAlex — the open scholarly corpus that replaces scraping for a great many bibliometric questions.
• Data collection methods — choosing a collection approach before choosing a vendor.
• Data management plan (DMP) — where your collection method and retention schedule belong.

Pricing checked against each vendor’s public pricing pages on 26 August 2026. Proxy list prices and promotional rates change frequently; verify current figures with the vendor before committing budget.