tokio vs tower-layer

Short answerBy steady daily use, tokio is bigger: 3,098,313 real downloads a day against 1,850,070 for tower-layer (1.7x). 78,796 crates depend on tokio and 342 on tower-layer. tokio last shipped 3 days ago, tower-layer 2.2 years ago.

tokio

Foundational

Foundational and steady.

78,796 crates depend on it; downloads jump on release days (1.4x the baseline).

An event-driven, non-blocking I/O platform for writing asynchronous I/O backed applications.

tower-layer

Established

Established and quiet for years.

342 crates depend on it; no release in the last 90 days, so its count is all steady use.

Decorates a `Service` to allow easy composition between `Service`s.

tokiotower-layer · 100% = that crate's normal quiet day, so a spike to 300% is triple its usual traffic

How they relate

Often used together: 5 tracked crates depend on both (tower, tower-http, hyper-util, axum, tonic). They are not always either/or.

Every number, side by side

Same window, same method	tokio	tower-layer
Real downloads a day	3,098,313	1,850,070
Crates that depend on it	78,796	342
Last release	1.54.1, 3 days ago	0.3.3, 2.2 years ago
Trend vs the field	-8%	+3%
Release-window share	11.6%	0.0%
Downloads in the window	272,815,137	172,795,738
All-time downloads	1,101,468,563	598,768,745
Versions published	201	7
License	MIT	MIT
MSRV	1.85	not set
First published	Jul 1, 2016	Apr 27, 2019

Which to pick

Downloads tell you what is common, not what fits. Each crate's own description above says what it is for; when the jobs differ, pick by the job.
When both fit, prefer the one more crates depend on (tokio, 78,796 dependents): more eyes, more answered questions, and you may already have it in your tree.
Check the release dates against your own pace: tokio 3 days ago, tower-layer 2.2 years ago.

Also in this space

futuresFoundationalAn implementation of futures and streams featuring zero allocations, composability, and iterator-like interfaces2,405,900 a day, 26,505 dependents atomic-wakerEstablishedA synchronization primitive for task wakeup1,911,010 a day, 116 dependents futures-liteWidely usedFutures, streams, and async I/O combinators763,457 a day, 1,109 dependents tonic-buildWidely usedCodegen module of `tonic` gRPC implementation683,184 a day, 1,000 dependents

Compare 4 of them side by side →

Method

Real downloads = the median day with no release that day or the day before, from the crates.io API (Jul 27, 2026 to Oct 23, 2026). Raw crates.io counts include CI and bots and run high (why); the method is from our 90-day study.

Up next

Another comparison: Same method, same window. Foundational and steady vs established and gaining.tokio vs smolRead it next →

Also worth your time