Why in-browser PDF tools are faster — measured
Merging two 6.28 MB documents in the browser took 113 milliseconds on our test machine. On the same machine, uploading one of those files at a measured 14 Mbps takes 3.6 seconds — before the server has done anything at all. The gap is not optimisation. It is the network.
Measured 27 July 2026 on an Apple Silicon Mac (10 cores, Chrome). Raw numbers and method below — you can repeat all of it.
What we measured
Test documents are image-per-page PDFs built with a distinct photographic-noise image on every page, so file size scales the way a real scan does. Operations ran in the live browser using the same engine the tools use. Upload throughput was measured by timing three 8 MB transfers to our own server — not estimated. It came out at 14 Mbps.
| Operation | In browser | Transfer alone |
|---|---|---|
| Merge two 6.28 MB PDFs → 12.56 MB | 113 ms | 4.8 s |
| Merge two 18.84 MB PDFs → 37.68 MB | 307 ms | 14.5 s |
The comparison is deliberately unfair — to us
The right-hand column is only the network transfer: upload the input, download the result. It assumes the server processes your file in zero time and that you never queue behind another user. That is a best case no real service achieves.
Even against that impossible standard, doing the work locally is roughly 43× faster on a 6 MB document and 47× faster on a 19 MB one. We did not measure any competitor's processing time, and we are not guessing at it — we simply compare against the part of their work that cannot be optimised away.
Why upload is the bottleneck
Home and mobile connections are asymmetric. You may have 200 Mbps down but only 20 up. Every upload-based tool pays that slow leg with your document, then pays again to return a result that is often larger than what you sent — a merge doubles it. Working in the browser removes both legs, which is why the advantage grows with file size rather than shrinking.
When a server wins
Honestly: on an old or low-powered device running something heavy — OCR across hundreds of pages, for instance — a powerful server CPU can beat local processing, and the upload may be worth it. There is also a one-off cost on your first visit while the engine downloads, though it is cached from then on. We would rather say this than pretend the trade-off does not exist.
Repeat it yourself
Open any of our tools, open DevTools, and use the Performance and Network tabs. You will see the work happen locally and no request carrying your file. The Privacy Lab covers the network side in detail.
Try it with your own file
Free, private, no account. The speed is a side effect of not uploading.