// BENCHMARKS

Measured on
local fixtures.

Honest numbers from the Koko microbench suite — not marketing claims. Results reflect active development on main; re-run benchmarks after major lifecycle changes.

// BENCHMARKS

Measured on
local fixtures.

Local fixtures only9:36:45 AM
0.00×
Navigation geomean vs Chromium
Local static fixtures (2026-06-29)
0ms
Cold start (Koko mean)
Comparable to bundled headless Chromium
0
Fixture pages benchmarked
dom-heavy, js-compute, minimal, mixed
0
Benchmark errors
Koko + Chromium, 3 repeats each
Navigation results (mean ms)
dom-heavy.html
19.9ms72.7ms0.27×
js-compute.html
8.1ms37.9ms0.21×
minimal.html
8.1ms37.9ms0.21×
mixed.html
9.4ms48.7ms0.19×

// METHODOLOGY

How these numbers are produced

Benchmarks are probe-driven engineering artifacts. Treat them as directional signals for automation infrastructure — not a guarantee for every production workload.

Fixture set

Four local HTML pages: dom-heavy, js-compute, minimal, and mixed. Static files only — no live network variance.

Comparison

Koko headless runtime vs Playwright-driven Chromium on the same machine, three repeats per page, mean reported.

Navigation metric

Time from navigation request to load event — the primary signal for automation cold-path performance.

Reproducibility

Run zig build microbench in the Koko repo. Full methodology and raw tables live in knowledge/ and the dev blog.

Deep dive: microbench baseline write-up on the engineering blog.

Want the raw tables?

Clone the Koko repository and run the microbench target locally. Compare against your own fixtures and hardware.