Research

Public PDF research and benchmark downloads

The numbers on this page come from a simulated test environment and do not represent real-world processing times. They compare the overhead of different We Want PDF tools under identical, artificial conditions. Each dataset is available as a CSV file with a short write-up.

Simulated test data

Direct-download PDF tool benchmark

Simulated test results for tools that return the finished file in the same request. Timings were measured against a stand-in processing service with a fixed delay, so they do not represent real-world processing times.

Simulated test data

Queued PDF job benchmark

Simulated test results for tools that run as background jobs. Timings were measured against a stand-in processing service with a fixed delay, so they do not represent real-world processing times.

How the tests were run

Both datasets were produced by sending small test requests to each tool through the website's normal request path. Instead of a real PDF engine, a stand-in service answered every request after a fixed delay. That keeps conditions identical across tools, but it also means no real document was processed.

Real processing time depends on your file: its size, page count, whether it contains scanned images, and the tool you use. A large scanned PDF sent to OCR will take far longer than any number shown here.

  • Each tool was run 3 times through the website's own request path in a simulated test environment.
  • The actual PDF processing step was replaced by a stand-in service that waits a fixed time (75 ms for direct-download tools, 250 ms for queued tools) instead of processing real documents.
  • Because file size, page count, and real processing work were not part of the test, the numbers only show relative overhead between tools, not how long a real document will take.

What the data can and cannot tell you

The files are useful for seeing how tools differ from each other under the same simulated conditions. They are not a measurement of real-world speed and should not be read as a performance claim or compared with other PDF services.

  • See which tools return a file immediately and which ones run as a queued job with a progress state.
  • Compare relative overhead between tools that were tested under the same simulated conditions.
  • Check the raw rows behind the summary in each companion article.

Each dataset has a short companion article that explains the setup and summarizes the pattern in the rows.

How to read the CSV files

The CSV files open in any spreadsheet app. Each row is one tool, with timings in milliseconds measured in the simulated test environment described above.

Because the processing step was simulated, the numbers should not be quoted as real-world performance data. If you refer to them, describe them as simulated test results and link to the matching article so readers can see the setup.

Columns in the files

  • slug: the tool that was tested, matching its /tools/ page.
  • iterations and fixed_upstream_delay_ms: how many runs were made and the fixed stand-in delay used in place of real processing.
  • Timing columns in milliseconds: average, median (p50), minimum, and maximum for direct-download tools; submit time and total time for queued tools.
Trust center

Trust pages

How We Want PDF handles your files: privacy, security, and how long uploads are kept.

privacy

Privacy

How We Want PDF handles the files you upload, what is needed to use the tools, and which tools send content to an AI provider.

Read page
security

Security

How We Want PDF checks uploads, delivers results, limits abuse, and removes queued job files.

Read page
retention

File retention

How long We Want PDF keeps uploaded files and results, and when they are removed.

Read page