FiretoolThese are measurements of Firetool 4.3.0 from its own test suite, not estimates. Every number below comes from one report (bench-4.3.0), and the test is repeated for each release.
The numbers
With 50,000 documents of 12 fields each, on the computer described below:
| What | Measured | Our limit |
|---|---|---|
| Start to a window you can use | 1.7 s (three starts: 1.5, 1.7 and 1.9 s) | 2 s |
| A query of 1,000 rows, drawn | 0.28 s (median of three runs) | 0.5 s |
| Load more of 1,000 rows, up to 50,000 loaded | Slowest 0.26 s; the last one 0.12 s | 0.4 s |
| Click or key to a visible response (sort, open a document, command palette, settings, filter) | Slowest 80 ms (the command palette); opening a document 55 ms, a sort 76 ms | 100 ms |
| Longest time the window was blocked | 100 ms | 200 ms |
| Memory when idle, all processes (private memory) | 287 MB | 350 MB |
| Memory per loaded row (12 fields) | 2.4 KB; 466 MB in all with 50,000 rows loaded | 8 KB |
| The window while an export runs | Slowest answer 26 ms | 200 ms |
| Export of 50,000 documents to CSV | 2.1 s (about 24,000 documents a second), 91 MB of memory, a 6.9 MB file | - |
| Backup of 50,020 documents | 3.1 s (about 16,000 documents a second), 50 MB of memory, a 1.7 MB file | - |
Every number is within its limit. The table view draws only the rows you can see, which is why 50,000 rows load without the window freezing, and exports and backups are written to the disk as they are read, so their memory stays small however many documents there are.
How they were measured
- Computer: a laptop with an Intel Core i5-10210U (8 threads), 16 GB of memory, Windows 10, on 5 October 2026, with other programs open. A newer or quieter computer will do better; runs on this one vary by about a third.
- Data: 50,000 documents with 12 fields (a third numbers, the rest text) in one collection.
- Test: the built app is started with a fresh profile and driven by an automated browser test (
tests/bench/bench.js). Memory is private memory of all Firetool processes, which is what Task Manager shows, not the larger working set. Click times are taken from the click to the change on screen; the longest block is the longest task the window's main thread ran. Backup and export are timed in the main process (tests/bench/jobs.js). - Limits: each number has a limit we hold ourselves to; a release that goes over one is looked at before it ships.
What they do not say
- The data came from a test server on the same computer, not from Firestore. A real query also waits for the network and for Firestore, so query, Load more, export and backup times with your project will be longer, and mostly set by your connection. The test server also slows down as the collection grows, so the Load more and query numbers are on the high side for Firetool's own work.
- They measure Firetool, not other tools. We do not compare with other products here.
- Reading documents is billed by Google: loading 50,000 rows reads 50,000 documents. Firetool shows the number of reads before a big job; see Tasks and cost estimates.
Related
- Backup and restore: whole-database backups that are written as they are read.
- Export: JSON, CSV and Excel exports of any size.
- Tasks and cost estimates: what a job reads before you start it.