BENCHMARKS · UPDATED 2026-07-13
Performance comparison.
jsonbolt performance compared to the tools you already use. Same machine, same files, same goal for every tool: open the file and let me search. The dataset is public, and the method is documented below.
cpu i9-13900HX
ram 32 GB 5200 MT/s
ssd NVMe PCIe Gen4
os Windows 11 23H2
build c7e9c78 · Jul 2026
1 GB · 20.3M records
employees_10-level_1-GB_formatted.jsonTOOL
TIME TO TREE
PEAK RAM
SEARCH, ALL MATCHES
jsonbolt
309 ms
1.04 GB
210 ms
Dadroit
793 ms
1.12 GB
597 ms
VS Code
16.9 s
crashes
not measured
Sublime Text
47.7 s
2.1 GB
not measured
200 MB · 3.9M records
employees_10-level_200-MB_formatted.jsonTOOL
TIME TO TREE
PEAK RAM
SEARCH, ALL MATCHES
jsonbolt
51 ms
233 MB
40 ms
Dadroit
162 ms
227 MB
138 ms
VS Code
3.04 s
3.4 GB
not measured
Sublime Text
8.72 s
435 MB
not measured
50 MB · 996K records
employees_10-level_50-MB_formatted.jsonTOOL
TIME TO TREE
PEAK RAM
SEARCH, ALL MATCHES
jsonbolt
16 ms
103 MB
14 ms
Dadroit
50 ms
61 MB
53 ms
VS Code
766 ms
1.4 GB
not measured
Sublime Text
2.08 s
122 MB
not measured
Methodology.
If a benchmark cannot be reproduced, it is marketing. Every cell above states exactly what was measured and how. Run the same files through the same tools, and email us if your numbers disagree with ours.
01
Warm runs. The first 3 runs are discarded as warm-up; each cell comes from the 5 runs that follow.
02
Time to usable tree. The clock stops when the tool accepts scroll and click input, not at first paint.
03
Search is exhaustive. Every match in the document, counted and navigable. Search was measured for the tools built around structured whole-document search (jsonbolt, Dadroit); plain-text find in editors is not comparable and is marked not measured.
04
Frame-counted timings. VS Code and Sublime Text times were captured with a high-speed screen recording and measured by frame comparison.
05
Built-in timing. jsonbolt and Dadroit times come from each tool's built-in timing output.
06
Public dataset. Employee files from sample.json-format.com, used as downloaded.
07
Default settings everywhere. Every tool at its latest stable version on the run date, out of the box.
Try it on your own files.
The best benchmark is the files that you deal with daily.