I ported your client-nav benchmark to our router — here are the results

[greydragon888]

Hi! I'm the author of https://github.com/greydragon888/real-router, a framework-agnostic router.
While working on performance, I ported your client-side navigation benchmark.
It exercises the same scenario against both TanStack Router React and Real-Router with React adapter under identical conditions.
I'm also planning to add benchmarks for Vue and Solid adapters soon — it'll be interesting to compare those as well.

Setup:

• Both routers use in-memory history (no window.history overhead)
• Each iteration runs 10 navigations: click-based + programmatic navigate + replace
• Same React 19 rendering pipeline, same route tree shape
• Benchmark powered by https://vitest.dev/guide/features.html#benchmarking
• Full source: https://github.com/greydragon888/real-router/tree/master/packages/router-benchmarks

Results (MacBook Pro M3 Pro, 18 GB RAM, macOS Sequoia):

Real-Router	TanStack	Real Router	Δ
hz (ops/sec)	109.73	58.40	1.88x быстрее
mean (ms)	9.11	17.12	1.88x быстрее
p75 (ms)	14.00	23.01	1.64x
p99 (ms)	19.11	40.29	2.11x
min (ms)	1.22	4.82	3.95x
rme	±3.37%	±4.21%	стабильнее
samples	1099	586	—

Observations:

The gap in minimum latency (1.2 ms vs 4.8 ms) suggests the difference is mostly in core overhead — route matching, state diffing, subscriber notification.
Not than in React rendering, which is shared.

Why share this here:

I think reproducible cross-router benchmarks are valuable for the ecosystem.
The benchmark repo is public and the setup is straightforward to run locally.
If something looks off methodologically, I'dappreciate the feedback. happy to adjust!