Independent Testing Cuts Floci's 24-Millisecond Claim to 3 Seconds

Floci, the open-source AWS emulator that filled the gap after LocalStack shut down its free Community Edition, advertises a 24-millisecond startup time in its own documentation, against LocalStack's own claimed 3.3 seconds. An independent benchmark published April 3 measured Floci itself at roughly 3 seconds instead, more than 100 times slower than the vendor's own figure.

What actually changed on March 23

LocalStack ended free support for its Community Edition on March 23, 2026, consolidating to a single Docker image that fails to start without an authentication token, the company said in a post co-written by co-chief executives Waldemar Hummer and Gerta Sheganaku. "We need a distribution model that lets us engage directly with users, understand how LocalStack is used, and sustainably invest in the platform," they wrote. The paid Base tier costs $39 per license per month, billed annually, according to Tech Times.

Developer Andrés Villarreal described the transition arriving as CI failures rather than an announcement. His pipelines, pinned to a latest LocalStack image, began "randomly failing in the middle of the day," he wrote on his blog. He traced the cause to the license change, pinned his build to version 4.4.0, the last free release, and said he started testing Floci as a replacement. His broader conclusion was about dependency hygiene rather than LocalStack specifically: "Always specify a deterministic version. Never use latest or similar," he wrote, extending the same advice to package managers and GitHub Actions.

LocalStack's Feb. 27 pricing post, updated July 24, kept a free tier alive for what it calls "non-commercial recreational usage only," added a 30-day free trial on paid plans, and removed CI credit limits across every tier. It also created an open-source sponsorship program that gives qualifying projects free access to paid packages, an option aimed at exactly the kind of maintainers who might otherwise migrate to a free alternative like Floci. A temporary environment-variable bypass that let existing pipelines keep running without the new token expired April 6, 2026, Tech Times reported, giving teams roughly two weeks past the hard cutoff to migrate before even the workaround stopped working.

Reaction split along predictable lines. Developer alvsanand, quoted by Tech Times, called it "ironic to read them calling it an 'open-source experiment' rather than a full project," referring to how LocalStack had described the Community Edition it was retiring. Hummer and Sheganaku's own post anticipated the criticism, framing the change as a way to keep the paid product funded rather than a retreat from open source.

Floci filled the gap before the deadline even arrived

Floci's own repository was created Feb. 18, 2026, five weeks before LocalStack's cutoff, under an MIT license and the tagline "The AWS Local Emulator alternative." It has since grown into a family of emulators, including floci-az for Azure and floci-gcp for GCP, and its main AWS repository has reached 25,688 stars and 2,779 forks, up from the 10,000 stars Tech Times counted in mid-May. Tech Times reported Floci covering 45 AWS services against LocalStack's broader catalog, packaged into a roughly 90-megabyte Docker image compared with LocalStack's roughly 1-gigabyte image, according to the same Quarkus issue that later tried and failed to add native Floci support.

The speed claims don't match independent testing

Floci's documentation claims roughly 24 milliseconds to start and about 13 mebibytes of idle memory, against LocalStack's own documented figures of about 3.3 seconds and 143 mebibytes. Both numbers are vendor-supplied, taken from each project's own materials rather than a shared test. Developer Nahuel Nucera ran an independent comparison on April 3, cold-starting each tool five times on an Apple Silicon M4 with Docker Desktop and taking the median, after a full docker system prune before every run.

Tool	Measured startup	Source
MiniStack	Under 2 seconds	Independent benchmark, April 3, 2026
Floci	About 3 seconds	Independent benchmark, April 3, 2026
LocalStack	15 to 30 seconds	Independent benchmark, April 3, 2026

Nucera's test also found Floci's memory use was not flat: about 26 megabytes at idle, rising to 56 megabytes after more than 500 operations, a pattern the post flagged as less predictable than MiniStack's roughly 38 to 39 megabytes across the same range.

Even a project with a ready-made reason to switch didn't

Quarkus, the Red Hat-backed Java framework, had a direct incentive to add Floci support: its dev-services tooling spins up local AWS emulators automatically for every developer who runs a test suite. A contributor opened a GitHub issue the same day LocalStack's cutoff took effect, proposing a quarkus.aws.devservices.provider=floci configuration option and citing Floci's no-auth startup and smaller image size. The issue was closed as not planned, with no public reasoning attached, leaving Quarkus's default dev-services path pointed at LocalStack even as Docker's own cloud sandboxes compete on price in the same developer-tooling budget line.

What's next

LocalStack has not published a benchmark of its own to counter Nucera's numbers, and Floci has not published a methodology to defend its 24-millisecond figure against them either. The dispute over emulator speed is a smaller-scale repeat of Nvidia's quiet move to the open-source Vault fork OpenBao after a commercial license change: a free alternative gains users during the argument, while the vendor-supplied numbers on both sides stay unverified until someone outside either company runs the test.