PDF Compression Benchmark
Last updated: August 27, 2026
PDFmuto's compression benchmark processed 8 public-domain documents totaling 6,059 pages through the in-browser Compress tool. Median file size fell 2.7%, five of eight files shrank, and the best result — an editor-saved 20.9 MB government PDF — compressed 31.1%. All processing ran locally; nothing was uploaded.
Key results at a glance
- Median reduction: 2.7% across 8 documents (6,059 pages).
- 5 of 8 files shrank. The other 3 grew by up to 7%.
- Best result: −31.1% on a 20.9 MB editor-saved government document.
- Total: −16.4% across the whole 51.3 MB corpus.
- Compression is a repack: it restructures the PDF without re-encoding images.
Methodology
We selected 8 public-domain and US-government documents that represent the way real PDFs behave: tax forms, legislative reference books, instructions with graphics, and a government gazette. The corpus spans 2 to 2,862 pages per file, from 137 KB to 20.9 MB.
Each file was processed with the exact settings the PDFmuto Compress tool applies in your browser:
- Load with metadata updates disabled (
updateMetadata: false), so editor metadata is preserved. - Producer and creator strings cleared, removing verbose strings injected by editors.
- Saved with object streams enabled (
useObjectStreams: true), the PDF format's compact encoding. - Run on 2026-08-27, in the browser's JavaScript engine, identical to a real user's run on this site.
The measurement is a single deterministic pass per file — no averaging over runs, no cherry-picking. Every source file is linked in the table below so anyone can reproduce the test.
Results
| Document | Pages | Before | After | Change |
|---|---|---|---|---|
| IRS Form W-4 (2024) | 5 | 204 KB | 208 KB | +2.2% |
| IRS Form 1040 (2023) | 2 | 215 KB | 217 KB | +0.8% |
| IRS Form W-9 (2024) | 6 | 138 KB | 131 KB | −4.5% |
| Instructions for Form 1040 | 126 | 4.2 MB | 4.5 MB | +7.0% |
| Declaration of Independence & Constitution | 85 | 3.8 MB | 3.8 MB | −0.9% |
| Constitution Annotated (2022) | 2780 | 13.4 MB | 12.6 MB | −5.7% |
| Constitution Annotated (2016) | 2862 | 20.0 MB | 13.8 MB | −31.1% |
| Federal Register, Jan 7 2022 | 193 | 7.0 MB | 5.7 MB | −18.5% |
| Total | 6059 | 48.9 MB | 40.9 MB | −16.4% |
Why some PDFs grow (and what that means for you)
The three files that grew — two IRS forms and the Form 1040 instructions — were already tightly packaged by the government's production pipeline. Repacking had nothing to remove, and the rewrite added a small amount of overhead. This is the honest trade-off of repack compression: it shines on documents saved by desktop editors, which tend to accumulate redundant data, and does little on professionally packaged files.
The best result, −31.1%, came from a 20.9 MB document saved by an older production tool. That pattern — large files, editor-generated, heavy metadata — is exactly where this type of compression wins. If your PDF was scanned or is dominated by images, repacking will not shrink it much; image re-encoding is the lever there.
Try it on your own file
The exact process measured here is the free PDF Compress tool. It runs in your browser with no upload — drop a PDF in, and the same pdf-lib pipeline produces your result on your device. For image-heavy documents, expect modest gains from repacking; for editor-saved documents, expect the kind of reduction the table above shows.
Curious how a website can process PDFs without a server? See ourlocal processing architecture explainer.
FAQ
How much can PDF compression actually shrink a file?
In this benchmark the median reduction was 2.7% and the mean 6.3% across 8 public-domain documents. Results range from 31.1% smaller (a bloated editor-saved document) to 7% larger (a file already tightly packaged). The outcome depends on how the PDF was created, not on the file's page count.
Why do some PDFs grow during compression?
Repack-style compression rewrites a PDF's internal structure but does not re-encode images. A file that is already tightly packaged — like the IRS forms in this benchmark — has little to gain, and the rewrite can add a small amount of overhead, up to 7% in the worst case here.
What is the difference between repacking and re-encoding a PDF?
Repacking removes redundant data and restructures a PDF without touching its images or text. Re-encoding recompresses images at a lower quality, which is where most size is won in image-heavy documents. The two methods are complementary; repacking alone cannot compress a scanned document much.
How was the benchmark run?
Eight public-domain US-government PDFs were processed with the same pdf-lib settings the live Compress tool uses on your device: metadata updates disabled, producer and creator strings cleared, and object streams enabled. Every source file is linked from the results table, so the test is fully reproducible.