Measure the engine on your hardware.
Transparent operations, generated corpora, and runtime context. No comparison theater.
RUN ON THIS DEVICE
Browser benchmark lab
No sample or result leaves the worker. Large runs allocate real memory; start small on mobile.
Choose a corpus size and run the same json-buf engine used by the workbench.
METHODOLOGY
What the numbers mean
Separate engine stages
Validation and node/depth profiling, SHA-256 metadata, and structural indexing are timed independently. Text encoding and worker transfer happen before the worker timer.
Warm browser process
The WebAssembly module is initialized before measurement. Each operation runs in the dedicated worker.
Generated, reproducible shape
A root object contains an array of consistent nested order records. Repository corpora add malformed, deep, numeric, and irregular cases.
Not a vendor leaderboard
Results vary with browser, CPU, thermal state, memory pressure, and extension activity. The lab reports your run.