Benchmarks

These results are without prefetch. A common misconception is that nitro-fetch is only faster when you prefetch — it isn't. Plain fetch is faster on its own; in our testing around 15–25% faster than React Native's built-in fetch for end-to-end requests (release build, device idle on Wi-Fi). Prefetch makes the difference far larger still, but it's not where the win comes from.

How much faster depends heavily on how your backend is set up — connection reuse, HTTP/2, and especially HTTP/3 over QUIC widen the gap.

Example 10-run head-to-head (14 public endpoints, alternating order, stall-corrected). "%" is how much lower nitro's latency was; "×" is built-in ÷ nitro; "Drop" is samples discarded for JS-thread stalls:

Run	Built-in (ms)	Nitro (ms)	% faster	×	Drop
1	196	152	22.5%	1.29	3
2	198	148	25.3%	1.34	1
3	197	148	24.9%	1.33	1
4	189	150	20.5%	1.26	2
5	189	146	22.6%	1.29	0
6	189	148	21.9%	1.28	3
7	187	147	21.4%	1.27	4
8	197	142	28.0%	1.39	4
9	190	144	24.5%	1.32	1
10	187	149	20.4%	1.26	3
Median	190	148	22.0%	1.28×	22 total
Average	192	147	23.2%	1.30×	—

Numbers are device-, network-, and time-specific. Try it yourself in the example app → the Benchmark V2 screen (or the Benchmark section in the tab navigator), on a release build.

If you have any concerns about the benchmark methodology, please open an issue.