Cloud Storage Radar

What Are the Best MinIO Alternatives for Local S3?

Published 2026-09-16 · Researched and price-checked on official sources

MinIO's open-source repository is archived, with no commits since 24 April 2026, so MinIO alternatives are now a live buying decision. Six replacements are worth shortlisting: Silo, SeaweedFS, RustFS, S3Proxy, Zenko CloudServer and Garage. Licenses, star counts and prices below were verified on 16 September 2026.

Key facts

What matters

The minio/minio repository is archived on GitHub: 61,366 stars, AGPL-3.0, and no push since 24 April 2026. What you run today is the last release that project will ship.
Silo, PGSTY's MinIO fork, is the closest drop-in for existing deployments: AGPL-3.0, 3,076 stars, commits landing daily, with packages, images, console builds and CVE fixes.
SeaweedFS (Apache-2.0, 34,719 stars) and RustFS (Apache-2.0, 32,315 stars) are the most active self-hosted options, but RustFS has 35 published security advisories, four of them critical.
Managed S3-compatible storage runs about $15.36 per TB-month on Cloudflare R2, $6.95 per TB-month on Backblaze B2 and $7.99 per TB-month on Wasabi, with no egress fees on R2 and Wasabi.
The license split decides commercial use: Apache-2.0 for S3Proxy, RustFS, SeaweedFS and Zenko CloudServer; AGPL-3.0 for MinIO, Silo and Garage.

Why MinIO alternatives are in demand right now

GitHub's API reports minio/minio as an archived repository: 61,366 stars, AGPL-3.0, no push since 24 April 2026. The company still sells object storage, but the freely redistributable server that DuckDB demos, Iceberg test harnesses and hundreds of Docker Compose files depended on is no longer where its engineering goes.

MinIO's own pricing page now lists AIStor in three tiers: a free single-node deployment for developers, researchers and small organisations; Enterprise Lite for horizontally scaled multi-node clusters with capacity below 400 TiB, price on request; and Enterprise with proactive support, price on request. A separate MemKV product targets GPU inference clusters.

None of that answers the practical question: what do you put in a Compose file when you need a local S3 endpoint this week, or a cheap S3 target you control?

What a MinIO replacement actually has to do

The January 2026 rmoff.net bake-off set five sensible requirements for local S3: a published container image, real S3 API compatibility, an OSI-approved license (Apache-2.0 preferred), single-node simplicity, and a maintainer still around next year.

That last point is not hypothetical. Of the projects below, only Apache Ozone sits under a foundation. Everything else is controlled by one company or one person, and MinIO itself relicensed from Apache-2.0 to AGPL-3.0 before archiving its repository.

Two checks pay for themselves before you commit: does the project publish security advisories, and does it have more than one significant committer? A single-maintainer project is fine until that maintainer stops.

The six MinIO alternatives, side by side

Everything in the table was read from the GitHub API on 16 September 2026, so the star counts and push dates are current rather than inherited from an older comparison. Star counts are a popularity signal only: they say nothing about release quality or security posture.

The star counts for SeaweedFS, RustFS, S3Proxy and Zenko CloudServer come from their primary repositories. Garage is developed on Deuxfleurs' own GitLab instance; the 4,522 figure is its GitHub mirror. MinIO is included as the baseline you may be replacing.

Silo: staying on MinIO without the archive banner

PGSTY forked MinIO in October 2025 and the fork is now published as Silo (pgsty/silo): AGPL-3.0, 3,076 stars, created 25 October 2025, with commits landing as recently as 16 September 2026. Its homepage lists server 20260903, console 2.4.0 and mcli 20260913, plus roughly 797,000 Docker pulls and a promise of packages, container images, a console and CVE fixes.

For teams already running MinIO, this is the lowest-effort path: the API surface, config format and client tools are the same, so migration is a container swap rather than a data migration. The caveats are the same as MinIO's were: AGPL-3.0 licensing, one organisation behind the project, and no foundation behind the governance. If the features MinIO removed from its community console were part of your workflow, confirm Silo's console covers them before you switch.

Licenses: where AGPL bites and where Apache does not

The self-hosted field splits cleanly. Apache-2.0 covers S3Proxy, RustFS, SeaweedFS and Zenko CloudServer. AGPL-3.0 covers MinIO, Silo and Garage.

That difference matters if you embed object storage in a product. Under AGPL-3.0, if you run a modified version as a network service, you have to offer the corresponding source to the people using that service. Apache-2.0 carries no such obligation, which is why vendors building storage into a commercial platform tend to start with SeaweedFS or Zenko CloudServer. Internal use on your own hardware does not trigger the AGPL requirement, so a self-hosted backup target on your own rack is normally unaffected.

If you would rather not run the storage yourself

Self-hosting only wins when the machine is already paid for and the data is replaceable. For everything else, managed S3-compatible storage is priced plainly enough to compare:

Cloudflare R2: $0.015 per GB-month for Standard storage, $0.01 per GB-month for Infrequent Access, with 10 GB-month free, no egress charges, Class A operations at $4.50 per million and Class B at $0.36 per million.
Backblaze B2: $6.95 per TB-month pay-as-you-go, free egress up to three times your stored volume.
Wasabi: $7.99 per TB-month pay-as-you-go, with no egress or API request fees, plus reserved-capacity pricing in one, three and five-year terms.

One terabyte of data costs about $15.36 a month on R2, $6.95 on B2 and $7.99 on Wasabi. A single-node self-hosted store only undercuts that if the server is idle anyway, which is usually the case on a budget VPS from a provider such as RackNerd, where the object store is a side job next to a database or a batch pipeline.

Which MinIO alternative should you pick?

Existing MinIO deployment, minimal change: Silo. Same API, same clients, maintained under AGPL-3.0.
New local test harness or CI: S3Proxy. Apache-2.0, a container image, and the least configuration of the group in the rmoff.net tests.
One node now, more later: SeaweedFS. Apache-2.0, the longest track record, and it scales out without a rewrite.
Multi-site cluster on your own hardware: Garage, if you accept AGPL-3.0 and a heavier initial configuration.
S3 in front of Azure, Google Cloud or another backend: Zenko CloudServer, backed by Scality.
No interest in running servers: Cloudflare R2 for cheap egress, Backblaze B2 or Wasabi for flat per-TB pricing.

Honest catches before you switch

RustFS is fast-moving and heavily patched. GitHub's advisory database lists 35 published advisories for the RustFS repository between 30 December 2025 and 8 September 2026, including four rated critical and 14 rated high, among them stored XSS leading to administrative account takeover (CVSS 9.0) and IAM privilege escalation (CVSS 8.8). That is a young, fast-remediated codebase: fine for a test bucket, a harder sell in front of customer data today.

S3Proxy has one maintainer and a retired dependency. The project is stable and Apache-2.0, but jclouds, one of the libraries it builds on, was moved to the Apache Attic in mid-2025. Local-file storage still works; the cloud-backend proxies are the part to test.

Garage is not a drop-in. It needs an extra initialisation container and a TOML config, and its key format is stricter than MinIO's. That strictness is good production hygiene and annoying for a weekend demo.

SeaweedFS reads like a commercial product. Its site leads with enterprise pricing while the project itself is Apache-2.0 with 764 open issues on GitHub. Check which edition a feature you need belongs to.

Archived is not broken. If MinIO is serving data today with no new requirements, you have time. The reason to move is a new CVE without an upstream patch, a feature you need, or a license review that flags AGPL.

At a glance

Tool	License	GitHub stars	Last push	Best for
Silo (MinIO fork, PGSTY)	AGPL-3.0	3,076	16 Sep 2026	Drop-in replacement for existing MinIO deployments
SeaweedFS	Apache-2.0	34,719	15 Sep 2026	Single node now, scale-out later
RustFS	Apache-2.0	32,315	16 Sep 2026	Greenfield projects that can track fast releases
S3Proxy	Apache-2.0	2,372	10 Sep 2026	Serving local files or other clouds as S3
Zenko CloudServer	Apache-2.0	1,947	15 Sep 2026	S3 front end to multiple backend clouds
Garage	AGPL-3.0	4,522 (mirror)	15 Sep 2026	Small multi-node clusters across sites
MinIO (reference)	AGPL-3.0	61,366	24 Apr 2026, archived	Workloads already running it

Licenses, stars and last push verified via the GitHub API on 16 September 2026.

FAQ

Is MinIO still maintained?

The open-source repository is not: minio/minio is archived with no push since 24 April 2026. MinIO the company still ships AIStor, whose free tier is a single-node deployment at no cost, with Lite and Enterprise subscriptions priced on request.

What is the easiest MinIO alternative to run in Docker?

In the January 2026 rmoff.net bake-off, S3Proxy and RustFS needed the fewest changes to an existing Compose stack, SeaweedFS and Zenko CloudServer were straightforward, and Garage required an extra initialisation container plus a TOML file.

Do I have to migrate off MinIO right now?

No. The last AGPL release keeps running, and PGSTY's fork at pgsty/silo tracks upstream with security fixes. Move when you need a new feature, a patch for a new CVE, or a license review that flags AGPL-3.0.