Comparison

This table compares VarMQ with ants, pond v2, conc, workerpool, errgroup and gopool. The cells are the same as in the feature comparison in the README. The notes marked VarMQ limit are new here: they say where a VarMQ Yes has a limit.

Find a feature

How to read it. Every row is a feature VarMQ has, so the VarMQ column reads Yes all the way down. The table does not list features that another library has and VarMQ lacks. Where a VarMQ Yes has a limit, the row has a VarMQ limit note, with the page of these docs that explains it. For speed, see the benchmarks.

Features of VarMQ, ants, pond, conc, workerpool, errgroup and gopool. Each cell says Yes or No.
Feature	varmq	ants	pond	conc	workerpool	errgroup	gopool
Generic (type-safe) API	Yes	Yes	Yes	Yes	No	No	No
Tasks return results and errorsVarMQ limit Only on in-memory queues. On a persistent or distributed queue, Add returns a bool and no job handle, so there is nothing to Wait on and no Err or Result to read. Worker types	Yes	No	Yes	Yes	No	No²	No
Task groups (submit many, wait once)VarMQ limit Only on in-memory queues. A persistent or distributed queue has no AddAll. Persistence overview	Yes	No	Yes	Yes	No	Yes	No
Priority queue	Yes	No	No	No	No	No	No
Persistent queues (survive restarts)VarMQ limit VarMQ has no adapter of its own: you add redisq (Redis) or sqliteq (SQLite). The best case is at-least-once delivery, so a job can run twice. What this guarantees	Yes	No	No	No	No	No	No
Distributed queues (across instances)VarMQ limit Of the existing adapters, only redisq (Redis) provides one. sqliteq has no distributed queue. An adapter of your own can implement IDistributedQueue. SQLite	Yes	No	No	No	No	No	No
Several queues bound to one pool	Yes	No	Yes⁴	No	No	No	No
Topic fan-out (publish once, many pools)VarMQ limit Topics are in-memory only. To share jobs between processes, use a distributed queue. Topics	Yes	No	No	No	No	No	No
Bounded queue, reject when fullVarMQ limit Not with redisq or sqliteq. Neither implements CapacitySetter, so Add is not refused on their persistent or distributed queues. An adapter of your own can implement it. Persistence overview	Yes	Yes	Yes	No	No	No³	No
Pause and resumeVarMQ limit On a persistent or distributed queue, Pause and Stop do not park a job the worker already took. It runs. Acknowledgement	Yes	No	No	No	Yes	No	No
Restart after stop	Yes	Yes	No	No	No	No	No
Resize pool at runtime	Yes	Yes	Yes	No	No	No	Yes
Cancel work via contextVarMQ limit Cancelling the context shuts the worker down. A running job is not interrupted: VarMQ cannot stop a function. Jobs that wait do not start. Context	Yes	No	Yes	Yes	No	Yes	No
Panic recovery	Yes	Yes	Yes	No¹	No	No	Yes
Task counters (submitted, failed, ...)	Yes	No	Yes	No	No	No	No
Built-in REST API	Yes	No	No	No	No	No	No

conc re-raises a task’s panic from Wait by design.
errgroup returns the first error, no results.
errgroup has no queue; TryGo only refuses a task while every slot is busy.
pond subpools: each has its own queue and runs on the parent pool’s workers.

Checked against each package’s source, October 2026.

What this page leaves out

This table covers in-process worker pools only. The benchmarks also measure Redis job queues (Asynq, gocraft/work, rmq, taskq) and SQLite job queues (goqite, Backlite, River). This page has no feature table for them.

For how fast each one runs, see the benchmarks, with the raw data in goptics/benchmarks, and the bench log.

What this page leaves out

On this page