This benchmark tries to compare different template engines in Rust. We consider both precompiled and dynamic engines, which of course have wildly different runtime performances.

Beware: One engine might be faster than the other. But if the other engine is more ergonomical to work with, e.g. because you have no re-compilation times, then it might still be the better fit for you. Only you can decide which characteristics of a template engine (runtime speed, simplicity of programming, …) are important to you.

Currently, we compare 17 different template engines. All of them are well maintained, mature and perform HTML escaping automatically.

Crate	Docs	Repo	Code	Recent Downloads	Github Stars	Contrib- utors	Recent Commits
askama			pre-compiled
dioxus			ssr
handlebars			interpreted
horrorshow			pre-compiled
hypertext			pre-compiled
leptos			ssr
liquid			interpreted
markup			pre-compiled
maud			pre-compiled
minijinja			interpreted
ructe			pre-compiled
sailfish			pre-compiled
sycamore			ssr
tera			interpreted
tinytemplate			interpreted
vy			pre-compiled
yew			ssr

Please see also Rust web framework comparison.

The benchmarking was done on Github's action runners, on an AMD-64 Linux system. The absolute performance can vary, but the relative performance of crate X to Y should be stable.

More information can be found in: https://askama-rs.github.io/template-benchmark/.

The "big table" benchmark generates a 100×100 cell HTML table. The "teams" benchmark contains a list of four sport teams and their scores. The former test contains only raw data, but a lot of it. The latter test includes texts that (might) need escaping, and "if" conditions.

To run the benchmarks on your machine, clone the repo, and execute cargo bench. The console will contain performance numbers, and the file target/criterion/report/index.html will contain more information. Run ./generate-table.py to generate the table you see above.