Rust vs. Go (Golang): Performance 2025

Go vs Rust Performance Comparison. Go (1.23.6) vs Rust (1.84.1).
🔴 To support my channel, I'd like to offer Mentorship/On-the-Job Support/Consulting ([e-mail])
🍿 Benchmarks: • New Benchmarks
👋 AWS is expensive - Infra Support Fund: buymeacoffee.c...
▬▬▬▬▬ Experience & Location 💼 ▬▬▬▬▬
► I’m a Senior Software Engineer at Juniper Networks (13+ years of experience)
► Located in San Francisco Bay Area, CA (US citizen)
▬▬▬▬▬▬ Connect with me 👋 ▬▬▬▬▬▬
► LinkedIn: / anton-putra
► Twitter/X: / antonvputra
► GitHub: github.com/ant...
► Email: [e-mail]
▬▬▬▬▬▬▬ Source Code 📚 ▬▬▬▬▬▬▬
► Original Source Code: github.com/ant...
#rust #golang #devops

[@AntonPutra]

🔴 To support my channel, I'd like to offer Mentorship/On-the-Job Support/Consulting ([e-mail]) 🍿 Benchmarks: kzbin.info/aero/PLiMWaCMwGJXmcDLvMQeORJ-j_jayKaLVn&si=p-UOaVM_6_SFx52H 👋 AWS is expensive - Infra Support Fund: buymeacoffee.com/antonputra

[@zuzelstein]

Now everybody is happy

[@valentinkaminskiy3826]

Add C++ 😂

[@ranggatohjaya5471]

Haha... me too😅

[@BlackistedGod]

there was a lot of crybabies on the previous test, some even said it should be deleted, although I love Rust, there's a lot of crybabies and cultist on its community

[@erikslorenz]

@@BlackistedGod The main use case is still fairly niche (versus like general backend development), so most Rust users are actual fans.

[@Jack-b4s3g]

Bravo!! Excellent work. You really listen the people. I see a great future for you and your channel.

[@AntonPutra]

thank you!

[@ОлегПахомов-ы3л]

Thanks for your work. 👍 Wait for c++ vs rust testing.

[@AntonPutra]

yes next

[@MaksimusPP]

It would be very interesting to compare userver c++ with rust.

[@Jack-b4s3g]

This would be VERY interesting to see !

[@ArchMaximus99]

Yeah we need c++ join the race 🎉

[@Mr.BinarySniper]

yeah. would be really awesome.

[@UnVFunLimited]

C++ wins

[@the-programing]

c++ built the world, rust built weirdos 😂

[@everyhandletaken]

Credit to Rust/Tokio/Hyper here, that performance is outstanding. I would still choose Go for my needs, but this is very useful to know, if ever that "next level" performance is needed. Thank you!

[@User948Z7Z-w7n]

You won't make a fast program until you gain deep understanding of memory alloc, concurrency, binary, and packaging. If you do, then you can absolutely scale Golang to the highest level. Cockroach DB is an example. If you don't, then no matter what language you learn, it will be mediocre

[@BlackistedGod]

@@User948Z7Z-w7n that only applies on few cases. if you tell newbies to create a web app in PHP (CodeIgniter/Lumen), Python (FastAPI), NodeJS (Express) and Go (Fiber). 99.9% of the time, GO will always be the fastest and PHP or Python will always be the slowest no matter how bad they are coded. and then tell the Senior's who have 100 years of experience in those language, Go would still win no matter how good they are coded.

[@donmorris4506]

Great work Anton👍

[@AntonPutra]

thanks!

[@Sam0sva1]

Oh how freaking awesome is Go from the box!

[@atajanov_kamronbek]

I'm super excited(impressed)!

[@vonderheimsk]

Thanks for this video! I would love to see Gleam vs Golang

[@МаксимМакаров-к8б]

Finally. It's quite fair enough.

[@AntonPutra]

yes

[@zakimimit]

I hope the next video: *Bun 1.2 vs rust (hono vs elysia vs fastify)

[@prakkumar3240]

why don't you compare the Drgon(c++) framework vs rust frameworks as these both languages are both system programming language and low level languages. Comparison of c++ and rust will be very intresting. Hope you read my comment😃😃

[@MMEEEish]

They're not low level

[@prakkumar3240]

@@MMEEEish according to me(as far as i know and learned) a low level language provide you access to memory and the hardware(both c++ and rust provide it). If you mean assembly is low-level and c++ & Rust are abstractions then you are completely true 👍.

[@AntonPutra]

i may do it next

[@MMEEEish]

@@prakkumar3240 it's a very common mistake. Even C is considered as a "mid level", because it has some abstractions. C++ and Rust are high level without garbage collection

[@armchair-architect]

Love these head to head comparisons. For me, at some point one has to also factor in time to market on a feature. At 60k requests per second when degradation starts, I would happily fire up another go container, and keep churning out features

[@svetlinzarev3453]

You assume that it's slower to develop in Rust, which is absolutely not true :)

[@samgarg5228]

@@svetlinzarev3453 It is indeed true. The language is simple. Standard lib has almost everything you need for web dev. There are more go devs (cheaper also). It is definitely easier to develop in Go as compared to Rust.

[@AnuragSingh-nx4ct]

Great work !!, my Channing Tatum.

[@EyeOfInfinity-t5g]

fiber is several times faster than the go standard library. It's strange that a third-party framework was chosen for rust, while a slow standard library was left for go.

[@smartmademarketing4259]

The issue with HTTP tests, unless you are using them as endpoints for some internal API, you usually need to add SSL, compression and connect to some sort of database which usually kills the some (if not most) of performance boost that using something like Hyper would give. Since Go gives better DX while still being fairly performant makes it so it usually ends up being my go (pun intended) to option

[@AntonPutra]

You offload TLS at the ingress/reverse proxy level, the app itself most of the time works with plain HTTP

[@smartmademarketing4259]

@@AntonPutra I'm a monolith type of guy, but I understand the need for distributed systems at big corporate level. The extra complexity is just something I avoid on personal projects.

[@ktappdev]

Rust boys smiling

[@omarmagdy1075]

Would be interesting to see Gleam vs Go

[@hasanaliyev5231]

Do the same with C++ vs Rust

[@AntonPutra]

yes next

[@nguyenmanh466]

The differences happened because of Go's garbage collector. Discord have also written a blog on how better the metrics were when they migrate their services from Go to Rust.

[@7777tiger]

It would be interesting to show how each of these scales with the number of cores. For example, 4, 8, 16, 32, 64, 128... Which one scales more linearly?

[@AntonPutra]

in real world you scale it on kubernetes by increasing number of replicas with 2, 4 or 6 CPUs

[@user-abc855]

Thank you for the video! It would be interesting to see Node.js with standard HTTP server vs Node.js with uWebSockets.js HTTP server vs Bun

[@AntonPutra]

ok thanks

[@jackking916]

impressive difference

[@PouriyaJamshidi]

Please give Nim a shot. It has a built-in async HTTP server as well.

[@eschoedler]

C# vs Rust? Thanks

[@Максим-д1ф4й]

Modern C++

[@AntonPutra]

ok

[@pink0ponk]

TOKIOOOOOO

[@onncpp]

Tokyo drift theme song

[@Quozul]

I developed a web server from scratch in C++, it may be faster than any other libraries, but also not production ready at all 😅

[@AntonPutra]

can you create PR and i'll test in a couple of days

[@jaaaj8516]

🎉🎉

[@AntonPutra]

:)

[@olhoTron]

Go still wins in the "lines of code per headache" metric

[@ibrahimshehuibrahim918]

I think go strength isnt in io bound operations,can you add real life CPU bound operations to your go vs rust test

[@AntonPutra]

will do!

[@advertentiegolf]

Can you also do a bun vs go test? Because it literally blew deno 2.0 (which is rust based) out of the water. I really want to know how close the performance of bun is to Go

[@daymaker_trading]

OMG!!!

[@AntonPutra]

yes

[@ОлегАндрианов-р7д]

Blazingly fast!) can you please measure rust with c++ userver framework?

[@AntonPutra]

yes

[@erikslorenz]

It's still crazy how performant Go (and stuff like C# and Java now) are while being GC general purpose languages that are pretty accessible. Rust/Zig is a nice step up if you are writing something more specific that requires serious throughput or power. The real loser with all this backend performance stuff imo is JS on the backend. Yeah it's faster than Ruby/Python/PHP, but there is nothing like Rails/Django/Laravel that give a full-stack experience or super easy experience like FastAPI. It lives in a weird no-mans land where the language isn't any easier than Go or C# or similar but you miss out on all the good stuff. Something like Next/Sveltekit could eventually become a real full stack framework but it really only exists to sell Vercel services, so no luck there.

[@samgarg5228]

Agreed but i guess javascript's main advantage is just because it's default on the web, frontend devs can easily shift to backend in js. You can code entire stack in a single language.

[@TariqSajid]

laravel octane vs go please please please !!!!!!!!!!!!!

[@warrenarnold]

Yea now we are talking

[@DavidSmith-ef4eh]

My prediction, 20% of go performance. so around 12k requests max

[@AntonPutra]

ok

[@MagicSyntax]

C++ vs Rust please

[@cangurcan99]

Hi what about a general TCP vs UDS (unix socket) benchmark? I believe will guide many devs.

[@winfle]

Still will use GO, since it is fast to develop, easy to scale in terms of concurency model, time to market much better. The only case when I will need Rust is when host cost is critical when it can cut bill for hundreeds of servers.

[@pauliusurmonas7665]

Good answer

[@svetlinzarev3453]

This is the way :) Where are the go shills now ? In the previous benchmark, the issue was that the rust toy server did not use keep-alive connections, while Go did. After implementing very rudimentary support for them, I hit a limit of ~60k rps for both Go and Rust, but Go was using 500% CPU, while Rust only 140% :) . I guess JMeter was simply not able to make more RPS as I was not able fully saturate my CPU cores, otherwise I'm pretty sure that Rust would have won even with the thread-per-request model due to to the significantly lower CPU usage :)

[@dmitryg6932]

When PHP runtimes? :)

[@AntonPutra]

1 week

[@channeldsr9983]

So, we have to see zig (maybe + lib) vs rust (tokio)

[@O...Maiden...O]

you can search "rust vs zig" on his channel

[@AntonPutra]

i'll refresh zig soon as well

[@jefriaritambunan2796]

go is the winner for short test but for long test rust is the winner of course

[@ЕвгенийКрасилов-о9о]

Rust vs ASSembly!

[@AntonPutra]

can you write assembly??

[@МаксимМакаров-к8б]

Наконец-то всё по-честному

[@AntonPutra]

haha

[@SuperRockmate]

For go use fiber is faster than default libraries Rust will still be faster just wanna know how much it will differ

[@AntonPutra]

it's based on fasthttp, but gnet is even faster

[@rednafi]

This is much fairer to Rust now. However, I don't enjoy Rust as a language and some of the large firms I've worked, Go is the default choice for writing RPC services and easier to hire for.

[@AntonPutra]

true

[@barkinarga4102]

Is tokio a framework for rust? And is it best one?

[@DavidSmith-ef4eh]

it's not a framework. I think it's a library that adds async support. And it's used widely inside rust development. Not just for servers, but for any scenario where you wan't asynchronous processing.

[@alexandrebeauchanps5992]

Tokio is an async runtime that as previously stated is widely used. Axum and Actix are the top dogs I believe in the ecosystem for now when it comes to servers. There are others that are relevant as well. If you look into the frontend framework, Leptos, Dioxus and Loco are in the making, but have yet to reach 1.0.0.

[@yurcchello]

why go isn`t using hyper?

[@Vaiz_]

please do a HAProxy vs nginx/apache2

[@AntonPutra]

ok

[@danyalhabib286]

Mojo vs Rust.... Please....

[@AntonPutra]

ok

[@xtraszone]

gleam vs node/bun vs go

[@rmnilin]

What about C/C++ tests? I think I can see the point of the video, but ultimately, it's a comparison between ASM and JavaScript. While a basic tutorial-level ASM web server might be less performant than even Next.js, there are endless possibilities to milk it for performance. This is an exaggeration, but I think the point is clear. On the other hand, comparing C/C++ top libraries to Hyper/Actix/etc. seems more relevant.

[@AntonPutra]

will do

[@Atmos41]

Can you do Rust Hyper vs C++ (for example Lithium or userver)?

[@AntonPutra]

yes maybe next

[@georgehelyar]

Can you just measure the memory of the processes?

[@AntonPutra]

Probably thanks, I'll see if Node Exporter exposes those metrics.

[@codispatch6869]

Rust + tokio = clear winner

[@AntonPutra]

yes

[@zhitoobit]

c vs rust

[@timmgrote2159]

@KZbin Stop auto translate shit without giving users a choice! We hate stupid AI auto translated voices, we hate auto translated titles, we hate auto translated descriptions! Commenting this under every video, until they change it.

[@atajanov_kamronbek]

fastapi only around+10k?

[@AntonPutra]

yeap sorry :)

[@Janwagame]

Your name look familiar. Are you Indonesian ?

[@AntonPutra]

nope

[@OlegKorsak]

so it seems that Python would lead to a faster time to market and then rewriting slow parts to Rust will bring the performance.

[@realharo]

I don't think Python would lead to a faster time to market (unless you already have an existing team with primarily Python experience, but that's true about every language)

[@theo-k4i8m]

@@realharo python should never be used to build web server

[@xura7CB]

I often see this argument. People try to rationalize the current market of technologies based on logical arguments. "If Python is slow, then it must be super productive, because otherwise why people use Python?" Or "performant languages must be super inconvenient to use, because that is why we have slow convenient languages" The main issue is that the popularity of the language is usually due to its history, not traits and qualities. Python is popular, because it was arguably the best and simplest dynamic language on the market. Also other technologies like native one (C/C++) and high-level (enterprise Java) were so annoying to use, so Python was the king of convenience I don't think we still live in this world. Native languages are much better, we as community care much more about performance and toolkit quality and progress in Python is just less visible than in other languages like Go or Rust.

[@realharo]

​@theo-k4i8m I wouldn't say that, I just think that productivity in most mainstream languages by skilled people can be so very similar that it almost doesn't matter what you choose

[@antifa_communist]

No. Rust is better in every way. "Rewrite the slow parts". So all of it.

[@atajanov_kamronbek]

Hi.

[@AilentRed]

How about bun vs Rust?

[@ranggatohjaya5471]

Bun made from zig, zig vs rust

[@DavidSmith-ef4eh]

he did bun the other day. Assuming he uses same servers, you can take the results from the bun vs deno vs node benchmars and compare them with those.

[@AntonPutra]

well, bun is zig wrapper

[@thinhbui4835]

Work with redis and db in rust is so hard compare to golang

[@youraccountissuspended]

Work with redis and db in golang is so hard compare to python

[@kevinvoyer6697]

Rust has nothing complicated ^^', it is just more verbose

[@123mrfarid]

​@@youraccountissuspendedworth it if it is for 40x performance according someone in medium

[@paco3447]

Try Erlang

[@ronaldosd]

Tokioooooooooooooooooooooooo (in primeagen voice)

[@AntonPutra]

haha

[@urparvezali]

there is the real winner.. axum would be more standard

[@AntonPutra]

true

[@elvispalace]

155k WTF?

[@AntonPutra]

crazy

[@dataquery]

C++ ?

[@AntonPutra]

soon

[@fabiopetrillo]

Rust is top!

[@AntonPutra]

yes

[@thibaud1213]

why tf do you move so much when talking

[@AntonPutra]

idk

[@winfle]

I'm sorry, did you take fastest library for Rust and standard library for Golang? Sounds unfair

[@svetlinzarev3453]

It's not the fastest library in Rust, but the de-facto standard HTTP server for rust. Essentially all rust servers build on top of hyper

[@damon.lee.r]

Golang 1.24 is due this month, from their pre-release notes: "Several performance improvements to the runtime have decreased CPU overheads by 2-3% on average across a suite of representative benchmarks. Results may vary by application"

[@ErenYagdran]

Rust is the king

[@infantfrontender6131]

As expected. Rust completely annihilated Go

[@hermes6910]

Go is still the best for perf/productivity, really the best of 2 worlds.

[@damon.lee.r]

@@infantfrontender6131 these are bare metal benchmarks. they're not representative of real world application use. once you factor processing and backend data stores, in the majority of cases, the gap will close, as they'll run as fast as the slowest subsystem.

[@svetlinzarev3453]

​@damon.lee.r nope, I've done that experimet, yku can find it in my github. Rust is around 2-5 times faster and uses 3 times less memory. I you have a target RPs, then rust achieves it with a lot less resources. Or if you have fixed resources - you get 5x their value with rust

[@AntonPutra]

yes

[@jabadabaduuuuuuuuuuuuu]

99% of people wathcing this will never run a server with 150 requests per second

[@warrenarnold]

Speak for yourself 😅

[@georgehelyar]

Yea one of the services I run in prod does about 6k requests per second with under 10ms response time while doing actual work, which ends up being about 500m requests a day, but we just scale out with docker containers, so even then each replica is not doing that, so all these micro benchmarks need to be taken with a pinch of salt, but it's nice to have a rough idea anyway.

[@jabadabaduuuuuuuuuuuuu]

@@georgehelyar can I ask what the service does?

[@BlackistedGod]

and why do you think everyone will start with $85 monthly server cost like the one being used in here? why would you spend $85 monthly if you're not hitting 150K req/s

[@jabadabaduuuuuuuuuuuuu]

@@warrenarnold Like I said, 99%, you are the 1%. What kind of service do you host? What causing all the traffic?

[@Shyam_Mahanta]

fun thing there is no RUST job.

[@oliverswanick4965]

skill issue

[@Shyam_Mahanta]

@oliverswanick4965 🥲

[@bcassol]

language issue

[@SDAravind]

Add zig to the mix

[@Anonymous-t3g]

Rust and Go both are OVERRATED and OVER-HYPE!

[@oboynitro]

exclaimed the ruby developer 😂😂😂

[@Anonymous-t3g]

@@oboynitro No, I know almost any major programming language. Rust is not a replacement of C/C++.

[@123mrfarid]

Tell me how they become popular then? Even money cant bought programmers loyality to language. Well money matters but if the programmers hate it then they wont use it.