Measurements and provider comparison
Below is a published measurement of knowmind. The corpus, the questions and the raw data are disclosed, so you can recalculate every figure. How well knowmind carries your own work is shown by a test with your own questions and documents.
Published measurement · 23 May 2026
Results you can trace to their source
The test ran 20 German questions that had not been used to tune the search, against 1,219 of our own memories. A hit means the expected source appears among the first five results. Tokens are the units of text an AI model processes and the provider bills for.
Scroll the table sideways to see all columns.
| Search method | Hit rate | Mean tokens sent to the model | Mean response time |
|---|---|---|---|
| Full-text search | 55 % | 3,244 | 26 ms |
| File-name search | 55 % | 2,660 | 2 ms |
| Vector search | 65 % | 1,600 | 491 ms |
| knowmind | 80 % | 1,805 | 475 ms |
A ranking against other providers only exists once everyone runs the same corpus, the same questions and the same scoring. Figures that each vendor collects for itself cannot be set against one another. The closest thing to your own case is a test with your own documents.