NewOut Plane raises funding, backed by Maza VenturesRead the announcement
Back to Blog
Comparison

MinIO Alternatives in 2026: 12 S3 Compatible Options Compared

Recep Erdoğan••14 min read
MinIO Alternatives in 2026: 12 S3 Compatible Options Compared

MinIO's own pricing page states that "MinIO OSS is at end of life and is no longer being maintained" and that "Its codebase is frozen, with no new features, bug fixes, or security patches." The repository was archived on 25 April 2026, and the last community binary shipped on 16 October 2025. If you self-host, the shortlist is SeaweedFS, Garage, Ceph with Rook or Apache Ozone. If you would rather buy it, look at Cloudflare R2, Backblaze B2, Hetzner, Scaleway, Tigris, Wasabi or Out Plane buckets.

Out Plane is our product. It appears once, in the managed section, with the same facts we checked for everyone else.

What Actually Changed at MinIO

Four dated events, each checkable on MinIO's own pages. All of the following was read on github.com/minio/minio and min.io/pricing on 15 November 2026.

24 May 2025: the console left the community build. The highlights of RELEASE.2025-05-24T17-08-30Z say the "Embedded UI Console is now deprecated and moved to object-browser", and that "External IDP logins via LDAP/OIDC are removed as well; these are now available as part of the AiStor Product." The same note adds that "STS APIs continue to work if someone wishes to build UI in front", and tells paying customers to upgrade to AIStor. That is the change most teams noticed first: the browser was still there, the administration screens were not.

16 October 2025: the last community binary. RELEASE.2025-10-15T17-29-55Z was published on 16 October 2025, and nothing has followed it. The README states that "The MinIO community edition is now distributed as source code only. We will no longer provide pre-compiled binary releases for the community version." Historical downloads stay up for reference, and the README is blunt about them: "These legacy binaries will not receive updates." The documented way to install the community edition is now go install github.com/minio/minio@latest, which needs Go 1.24 or newer.

25 April 2026: the repository was archived. GitHub shows "This repository was archived by the owner on Apr 25, 2026. It is now read-only", with the last commit on 24 April 2026, and the README note at the top points to AIStor Free and AIStor Enterprise instead. The command line client, minio/mc, is archived too, with its last commit on 20 November 2025. The Go SDK, minio/minio-go, is the exception: it is Apache-2.0, still open, and was last pushed on 12 November 2026.

Now: AIStor is the product. The AIStor pricing page says "MinIO AIStor is available in three tiers: AIStor Free, AIStor Enterprise Lite, and AIStor Enterprise" and that "MinIO AIStor is commercially licensed." AIStor Free is "Free of charge" with "Self-service community Slack and documentation support", but it is a "Single-node deployment architecture" and "Redistribution is prohibited." Enterprise Lite targets capacity "below 400 TiB", and Enterprise is quoted on usable capacity with no list price.

The licence of the old code did not change. The server is still AGPL v3, so the archived code remains legally forkable, which is why community forks exist. What it no longer has is anyone shipping security patches for it.

Two practical consequences follow. If you ship MinIO inside your own product, AIStor Free is not an option, because redistribution is prohibited. If you run a distributed cluster today, the maintained vendor path is a commercial contract.

MinIO Alternatives at a Glance

The first two rows are MinIO itself, so you can see what you are comparing against.

OptionTypeLicence or modelPrice on 15 November 2026EgressChoose it if
MinIO communitySelf-hostedAGPL-3.0, repository archived 25 Apr 2026Your hardware, source build onlyYoursYou already run it and will build from the last commit
MinIO AIStor FreeVendor buildCommercial licence, redistribution prohibitedFree of charge, single nodeYoursYou want MinIO's own maintained binary on one node
SeaweedFSSelf-hostedApache-2.0Your hardwareYoursBillions of small files, with an S3 gateway in the same binary
GarageSelf-hostedAGPL-3.0Your hardware, 1 GB RAM minimumYoursThree sites on ordinary internet links and modest hardware
Ceph with RookSelf-hosted on KubernetesCeph LGPL-2.1 or LGPL-3.0, Rook Apache-2.0Your hardware, 3 OSDs on 3 nodes in the sampleYoursYou already run Kubernetes and want block, file and object together
Apache OzoneSelf-hostedApache-2.0, version 2.2.2Your hardwareYoursLakehouse, analytics and AI pipelines over large files
Zenko CloudServerSelf-hostedApache-2.0Your hardwareYoursYou want an S3 front end over storage that lives on other clouds
Cloudflare R2ManagedCommercial service$0.015 per GB-month Standard"no charges for egress bandwidth for any storage class"Free egress, and Workers in the same account
Backblaze B2ManagedCommercial service$6.95 per TB-monthFree up to 3x average monthly storage, then $0.01 per GBCheap bulk storage, backups and archives
Hetzner Object StorageManagedCommercial service€6.49 per month excl. VAT, including 1 TB storage and 1 TB egress1 TB included, then €1.0000 per TB excl. VATBuckets beside Hetzner servers in Germany or Finland
Scaleway Object StorageManagedCommercial service€0.01606 per GB-month Standard Multi-AZ, Paris75 GB free every month, then €0.01 per GBYou must name the EU region, and you want object lock
TigrisManagedCommercial service$0.02 per GB-month StandardFree, including region to region and internetObjects that follow access patterns worldwide, with no egress bill
WasabiManagedCommercial serviceFrom $7.99 per TB-month pay as you go"No fees for egress or API requests"Flat per TB pricing, and 90 day minimum retention suits you
Out Plane bucketsManagedCommercial serviceInside the plan fee: Starter $9 per month, Pro $29, both with a 14 day free trialFree, unmeteredBuckets next to your apps, PostgreSQL and volumes in one console

Facts checked on 15 November 2026 against each project's or vendor's own pages: the MinIO repository, MinIO release notes and AIStor pricing, the SeaweedFS repository, Garage, the Ceph repository and Rook object storage docs, the Apache Ozone docs, Zenko CloudServer, R2 pricing, B2 pricing, Hetzner Object Storage, Scaleway storage pricing, Tigris pricing, Wasabi pricing and the Out Plane bucket docs. Licences come from each project's own repository. Prices change, so confirm them before you commit.

How to Pick in Two Questions

Does the data have to sit on hardware you control? If a contract, an air gap or a 200 TB archive says yes, you want the self-hosted list and you keep owning upgrades, disks and monitoring. If not, a managed bucket removes the part of MinIO that actually cost you time, which was never the S3 API. Our guide on self-hosted versus managed works through that trade in general terms.

Are you serving reads to the public internet? This is where the money is. Egress, not storage, dominates most object storage bills. R2, Tigris and Wasabi charge nothing for it, B2 gives you 3x your stored volume for free, Scaleway gives 75 GB a month, and Hetzner includes 1 TB in the base fee. Price your real monthly transfer before you compare per GB storage rates.

Self-Hosted MinIO Alternatives

These keep the model you have now: your disks, your failure domains, your upgrades. If you want them on managed infrastructure rather than a VM you patch, running self-hosted apps in Docker covers the general approach, and Docker hosting covers where the container runs.

1. SeaweedFS: built for very large object counts

SeaweedFS is Apache-2.0 and the most active project on this list, at 36,076 stars with commits on 15 November 2026. Its own description calls it a distributed storage system for object storage, file systems, Iceberg table buckets and Lance vector buckets, "designed to handle billions of files with O(1) disk access and effortless horizontal scaling." That small-file focus is the real differentiator: the design is volume files plus a filer, not one file per object.

The S3 layer is a gateway over the filer. The project's S3 API page documents three ways to start it: "weed s3 starts a stateless gateway server to bridge the Amazon S3 API to SeaweedFS Filer", weed server -s3 starts a master, a volume server, a filer and the gateway together, and weed filer -s3 starts a filer and the gateway. Coverage is broad, including versioning and multipart uploads. The page marks a list of calls unsupported, among them RestoreObject, SelectObjectContent, PutBucketLogging, GetBucketReplication and GetBucketWebsite, and notes that lifecycle "Transition rules not supported", MFA Delete is absent, and a delimiter other than "/" is not supported.

Operationally you now run several roles rather than one process: master, volume servers, filer and gateway, plus a metadata store for the filer at scale. That is the cost of the small-file design.

Choose it if you have tens of millions of objects or more, or you want file and object access to the same data. Look elsewhere if you want a single process on one node and nothing to think about.

2. Garage: the closest thing to a small, boring S3

Garage is AGPL-3.0, and its site frames it exactly as the gap MinIO left: "An S3 object store so reliable you can run it outside datacenters." It ships as a single binary with no dependencies, and the stated requirements are small: any x86_64 CPU from the last ten years, ARMv7 or ARMv8, 1 GB of RAM, and at least 16 GB of disk.

The design goals page is unusually honest about what it is not. "Extreme performances" and "Feature extensiveness" are listed as non-goals, and erasure coding is rejected on purpose, because "erasure coding or any other coding technique both increase the difficulty of placing data and synchronizing", so Garage limits itself to duplication. Each chunk of data is replicated in 3 zones. It is made for multi-site setups "interconnected through regular Internet connections", which is the opposite of Ceph's assumption of a fat local network. The documented deployments are Deuxfleurs, with 9 nodes across 3 locations, and Triplebit, with 15 storage nodes across 3 locations.

Choose it if you have two or three small sites, modest hardware, and you want an S3 endpoint that stays up without a storage team. Look elsewhere if you need raw throughput, erasure coding, or a long feature list.

3. Ceph with Rook: object storage as one output of a storage platform

Ceph is the heavyweight. Its README states that "Most of Ceph is dual-licensed under the LGPL version 2.1 or 3.0." The object interface is the RADOS Gateway, and Rook, Apache-2.0, is the Kubernetes operator that makes it practical: you create a CephObjectStore, and the docs say "the Rook operator will then create all the pools and other resources necessary to start the service", after which RGW pods serve S3 and Swift.

The shape of a production cluster is visible in the docs. The sample replicates metadata pools at size 3 and erasure codes the data pool at 2 data chunks plus 1 coding chunk, and it warns that "This sample requires at least 3 OSDs, with each OSD located on a different node." For real deployments it recommends "more chunks, such as 4+2 or 8+4." Access is Kubernetes-native: a CephObjectStoreUser produces access and secret keys in a Secret, and buckets come from a StorageClass plus an Object Bucket Claim, where "When the OBC is created, the Rook bucket provisioner will create a new bucket."

That is the honest trade. You get block, file and object from one cluster, erasure coding, and a project that is not going anywhere. You also get the largest operational surface on this list, and you should have someone who wants to own it. If you are weighing up Kubernetes tooling generally, container management tools covers the layer above.

Choose it if you already run Kubernetes at scale and want one storage platform for everything. Look elsewhere if object storage is the only thing you need.

4. Apache Ozone: object storage for analytics

Apache Ozone is Apache-2.0, and the current documented version is 2.2.2. Its docs describe an Ozone Manager and a Storage Container Manager that "handle metadata operations" with Datanodes that "manage the physical storage of data blocks", plus an S3 compatible REST API so that "applications developed for S3 to run on-premises using Ozone."

The positioning is specific and worth taking at face value. Ozone says it is "designed for lakehouse workloads, AI/ML, and cloud-native applications" and is "Intended for streaming reads and writes of large files." If your MinIO cluster backs Spark, Trino, Iceberg or a training pipeline over large Parquet files, this is the most natural fit on the list. If it backs user avatars, it is the wrong tool.

Choose it if your buckets feed a data lake or analytics engine and your objects are large. Look elsewhere if you store many small files or want something a single engineer can run.

5. Zenko CloudServer: an S3 API in front of other people's storage

Scality publishes two related Apache-2.0 projects. CloudServer is described in its own repository as "an open-source Node.js implementation of the Amazon S3 protocol on the front-end and backend storage capabilities to multiple clouds, including Azure and Google." Zenko is the larger product around it, "the open source multi-cloud data controller: own and keep control of your data on any cloud."

This solves a different problem from the rest. It is not primarily a place to keep bytes on your own disks; it is an S3 endpoint and replication layer in front of storage that lives elsewhere. If what you liked about MinIO was presenting one S3 API over mixed backends, this is the direct replacement.

Choose it if you need one S3 endpoint over several clouds, with replication between them. Look elsewhere if you want a storage engine that owns your disks.

Also considered

  • versitygw. Apache-2.0, 3,128 stars, last pushed 13 November 2026, and its repository calls it "A simple to deploy but feature rich S3 object storage server for your filesystem." If your MinIO deployment was really one node exposing an S3 API over a mounted disk, this is a closer match than any clustered system here.
  • MinIO AIStor Free. Still MinIO, still maintained by MinIO, free of charge, but single node, support via community Slack only, and redistribution prohibited. Worth knowing about if you only need one node and you are not shipping it to customers.
  • A community fork of the archived code. The AGPL v3 licence permits it, and the README invites it. The question to ask is who is reviewing security patches, because MinIO is not.

Managed S3 Compatible Alternatives

All prices below were read on each vendor's own pricing page on 15 November 2026.

6. Cloudflare R2: free egress as the headline

R2 lists Standard storage at $0.015 per GB-month, Class A operations at $4.50 per million and Class B at $0.36 per million. Infrequent Access is $0.01 per GB-month with Class A at $9.00 per million, Class B at $0.90 per million, and $0.01 per GB of data retrieval. On egress the page is unambiguous: "There are no charges for egress bandwidth for any storage class." The monthly free tier is 10 GB-month of storage, 1 million Class A and 10 million Class B requests, and it "only applies to Standard storage, and does not apply to Infrequent Access storage."

Choose it if you serve a lot of public reads, or your app already runs on Workers. Look elsewhere if you need to name the country your objects sit in on a price list.

7. Backblaze B2: the cheapest per terabyte

B2 lists storage at "$6.95 / TB / mo", billed per byte-hour, with the first 10 GB always free. Egress is free up to 3x your average monthly storage, with overage at $0.01 per GB, and egress above that multiple is also free through listed CDN and compute partners. Class A, B and C API calls are free on pay as you go; Class D calls cost $0.004 per 10,000 with the first 2,500 a day free. There are no minimum file size and no minimum storage duration fees.

Choose it if you are storing backups, archives or media and want the lowest storage line on the invoice. Look elsewhere if your read volume regularly exceeds 3x what you store and you are not behind a partner CDN.

8. Hetzner Object Storage: a flat fee with a quota

Hetzner prices Object Storage as a base fee rather than per GB: €6.49 per month excluding VAT, billed at €0.0104 per hour with that monthly cap, and the fee includes 1 TB of storage (up to 744 TB-hours) and 1 TB of egress. The quota is shared across all your buckets and does not roll over. Beyond it, storage is €0.0087 per TB-hour excluding VAT, and the page lists extra egress at €1.0000 per TB excluding VAT. Inbound traffic, internal traffic inside the eu-central network zone, and PUT, GET and DELETE operations are free.

The docs list three locations, Falkenstein (fsn1), Nuremberg (nbg1) and Helsinki (hel1), and some firm limits worth reading before you migrate: up to 100 TB and 50,000,000 objects per bucket, up to 750 requests per second, 5 TB per object with 5 GB per single PUT, a minimum billable object size of 64 kB, and stored objects are immutable, so the service is meant for write once, read many use.

Choose it if your servers are already at Hetzner and roughly 1 TB covers you. Look elsewhere if you need to overwrite objects in place, or you expect more than 750 requests per second per bucket.

9. Scaleway Object Storage: named EU regions and object lock

Scaleway's storage pricing page, read for the Paris region, lists Standard Multi-AZ at €0.000022 per GB-hour, roughly €0.01606 per GB-month, Standard One Zone at €0.000011 per GB-hour, roughly €0.00803 per GB-month, and Glacier at €0.0000034 per GB-hour, roughly €0.00246 per GB-month. Requests and ingress are included. Egress is 75 GB free every month then €0.01 per GB, and restoring from Glacier costs €0.009 per GB.

The concepts page lists four regions with their endpoints: fr-par in Paris, nl-ams in Amsterdam, pl-waw in Warsaw and it-mil in Milan. Standard Multi-AZ is available in Paris, Amsterdam and Warsaw, Standard One Zone in all regions, and Glacier in fr-par and nl-ams. Object lock is documented with Compliance and Governance retention modes and legal holds, which matters if you are replacing a MinIO cluster that held WORM data.

Choose it if you need to point at a specific EU region, or you need object lock and lifecycle rules. Look elsewhere if heavy public egress would blow past the 75 GB allowance.

10. Tigris: global placement with no egress charge

Tigris lists Standard storage at $0.02 per GB-month for Global and Single-Region, $0.025 for Multi-Region and $0.02 per region for Dual-Region. Infrequent Access is $0.01 per GB-month with a 30 day minimum retention and $0.01 per GB retrieval, Archive is $0.004 per GB-month with a 90 day minimum, and Archive Instant Retrieval is $0.004 with $0.03 per GB retrieval. Class A requests cost $0.005 per 1,000, Class B $0.0005 per 1,000, and DELETE and CANCEL are free. Egress is free, and the page states that Tigris "does not charge for regional data transfer", covering region to region and internet egress. Every account gets 5 GB of Standard storage, 10,000 Class A and 100,000 Class B requests a month.

Its Global mode is the distinctive bit: "Data distributed globally, follows access patterns", billed once with no per region charge. Dual-Region lets you name specific regions when you need to control where copies live.

Choose it if your users are spread across continents and you do not want to think about placement. Look elsewhere if $0.02 per GB-month is more than your storage budget allows at volume.

11. Wasabi: one flat rate, with conditions

Wasabi's pay as you go tier starts at "$7.99 TB/month", billed per GB per day for actual usage, and the page states "No fees for egress or API requests." Two conditions shape whether that is cheap for you: the minimum storage duration is 90 days, and there is a minimum monthly charge tied to 1 TB of active storage per account. Reserved Capacity Storage starts at 25 TB on 1, 3 or 5 year terms with a 30 day minimum duration.

Choose it if you store at least a terabyte of long-lived data and want one predictable per TB number. Look elsewhere if your objects are short-lived, since you pay for 90 days regardless, or you store well under 1 TB.

12. Out Plane buckets: S3 storage next to the app that writes to it

Out Plane runs S3 compatible buckets in the same console as the application, so the service that uploads files and the bucket it uploads to are configured in one place. You create a bucket with a name of 3 to 38 characters, lowercase letters, numbers and single hyphens, and choose public or private. The docs are precise about what those mean: "A bucket is private unless you make it public, and it has no access key until you create one", a private bucket means "Only requests signed with an access key can read files", and for a public one "Anyone with a file's link can read it. Listing stays off."

Access keys are scoped to a single bucket, and "A key works on this bucket only." A key is either read and write, which "can read, upload and delete files", or read only, which "can read and list files." The connection block is ordinary S3:

bash
AWS_ENDPOINT_URL_S3=https://obj.outplane.dev
AWS_REGION=auto
AWS_ACCESS_KEY_ID=<access-key-id>
AWS_SECRET_ACCESS_KEY=<secret-access-key>
BUCKET_NAME=<bucket-name>

Two details to set in your client before you migrate. A bucket's region shows as Global, and S3 clients sign with the region auto. Addressing is virtual-hosted only, so path-style requests do not work. Documented limits are 5 buckets per team on every plan, a largest file of 5 TB, and presigned URLs valid for up to 90 days. A public bucket serves files from https://<bucket-name>.obj.outplane.dev/<file>.

Pricing on 15 November 2026. Buckets are part of the plan rather than a separate per GB line. Starter is $9 a month plus usage with $9 of usage included, Pro is $29 with $15 included, and Growth is $299 with $150 included. Starter and Pro both start with a 14 day free trial: "Free for 14 days. Cancel before it ends and you pay nothing." The pricing page states that bandwidth, SSL and custom domains are not billed, and describes bandwidth as "Free, unmetered". Compute is metered by the minute while an instance runs. Custom domains come with TLS issued automatically once the DNS record resolves. Out Plane is GDPR compliant, and the outplane CLI is open source under Apache-2.0. Current figures are on the pricing page.

Choose it if your buckets exist to serve an application you are already deploying, and you want keys, apps, PostgreSQL and volumes in one console with egress you do not pay for.

Running an Object Store Yourself, on Managed Infrastructure

There is a middle path between a bucket you rent and a cluster you rack. You can run SeaweedFS, Garage or MinIO itself as an ordinary container on a platform that handles the host, the TLS and the deploys, and keep full control of the storage software.

On Out Plane that means a one-click template, 42 of them, including MinIO, attached to a persistent volume. You own the configuration, the version and the upgrade schedule; the platform owns the machine. It is the same pattern teams use for a self-hosted Redis instance. Applications deployed from a Git push or a Docker image run in Germany, with managed PostgreSQL through Neon in 8 regions and Frankfurt as the default alongside them.

This suits a team that left MinIO over the binaries rather than over the operational work, and still wants one S3 endpoint it controls.

Migrating Off MinIO

The good news is that almost nothing about your application changes. Every option above speaks S3, so the move is usually an endpoint, a region string, a key pair and an addressing mode. The work is in the data and in the things that are not objects.

Before you copy anything

Write down four numbers from the source: total size, object count, largest object, and how much of it is actually still read. Many MinIO clusters turn out to hold a working set of a few gigabytes under terabytes of old build artefacts, and the cheapest migration is the one where you delete first.

Then list what will not travel with a copy tool:

  • Bucket policies and IAM. MinIO users, groups and policies are MinIO's own. Every target has a different access model, so plan to recreate them. On Out Plane, for example, the equivalent is an access key per bucket with read and write or read only.
  • Versions. A copy tool reads the current version of each object. Version history is not carried across unless you deal with it explicitly.
  • Lifecycle rules, object lock and retention. These are bucket configuration, not objects. Check your target supports what you rely on before you start, because not every option on this list does.
  • Hardcoded endpoints and presigned URLs. Grep for your old hostname in application code, infrastructure manifests, CDN configuration and database columns. Previously issued presigned URLs will stop working when the old endpoint goes away.

Option A: rclone, for moving between two different providers

rclone is MIT licensed and its S3 backend documents MinIO as a provider, alongside Cloudflare, Scaleway, Hetzner, Wasabi, SeaweedFS and a generic Other, described as "Any other S3 compatible provider." This is the right tool for a cross-provider move.

Run rclone config, choose n) New remote, pick s3, and answer the prompts for provider, credentials, region and endpoint. The docs note that endpoint is "Required when using an S3 clone", and for region to "Leave blank if you are using an S3 clone and you don't have a region." You end up with two sections in your config file:

ini
[minio]
type = s3
provider = Minio
access_key_id = OLD_ACCESS_KEY
secret_access_key = OLD_SECRET_KEY
endpoint = https://minio.internal.example.com
region =

[target]
type = s3
provider = Other
access_key_id = NEW_ACCESS_KEY
secret_access_key = NEW_SECRET_KEY
endpoint = https://obj.outplane.dev
region = auto

Copy once, then sync until the delta is small:

bash
rclone copy minio:photos target:photos
rclone sync --transfers 200 --checkers 200 --checksum minio:photos target:photos
rclone check minio:photos target:photos

Three things from the docs are worth internalising. First, --checksum is the flag to reach for here: "If the source and destination are both S3 this is the recommended flag to use for maximum efficiency." Second, the parallelism matters, and the docs give 200 checkers and 200 transfers as an example for large S3 to S3 work. Third, and this is the one that surprises people: "For rclone to use server-side copy, you must use the same remote for the source and destination." Across two providers, every byte travels through the machine running rclone. Run it on a host with good connectivity to one end, not on your laptop over a home connection.

Option B: mc mirror, for MinIO to MinIO

MinIO's own client still works, with the caveat that minio/mc is archived and AGPL-3.0, its last commit was 20 November 2025, and its reference documentation now lives under the AIStor enterprise docs. If you are moving from one MinIO deployment to another, or to AIStor, it is still the most direct path.

Register both ends. The documented syntax is mc [GLOBALFLAGS] alias set ALIAS URL ACCESSKEY SECRETKEY:

bash
mc alias set old https://minio.internal.example.com OLD_ACCESS_KEY OLD_SECRET_KEY
mc alias set new https://new-store.example.com NEW_ACCESS_KEY NEW_SECRET_KEY

Then mirror, whose syntax is mc [GLOBALFLAGS] mirror [FLAGS] SOURCE TARGET:

bash
mc mirror --dry-run old/photos new/photos
mc mirror --overwrite --preserve old/photos new/photos

The flags that decide the outcome, in the docs' own words: --overwrite "Overwrites object(s) on the TARGET", and without it "if an object already exists on the Destination, the mirror process fails to synchronize that object"; --remove "Removes object(s) on the Target that do not exist on the Source", which is how you make the two sides identical; --preserve will "Preserve file system attributes and bucket policy rules of the SOURCE on the TARGET", which is the one real advantage over rclone here; and --watch "continuously synchronizes files from Source to Target until explicitly terminated", which is how you keep the target warm during a cutover.

The cutover

Mirror the bulk while the old store is still serving. Start a --watch run, or a short rclone sync loop, so the delta stays close to zero. Then put the application in read only mode or take a short write pause, run one final sync with --remove or rclone sync, flip the endpoint, region and credentials in your configuration, and redeploy. Verify with a read and a write against the new bucket before you let writes back in. Keep the old store intact and readable for at least a week; it is cheap insurance, and it is the only thing that makes the rollback trivial.

Frequently Asked Questions

Is MinIO dead?

The open source server is no longer developed. MinIO's pricing page says "MinIO OSS is at end of life and is no longer being maintained" and that its codebase is "frozen, with no new features, bug fixes, or security patches." The repository was archived on 25 April 2026 and the README says "THIS REPOSITORY IS NO LONGER MAINTAINED." The code is still there and still AGPL v3, so it runs and can be forked, but nobody is shipping security patches for it. MinIO the company is very much alive, selling AIStor in three commercial tiers.

Is MinIO still free and open source?

Both answers are partly yes and the details matter. The archived community server is open source under AGPL v3, and you can build it with go install github.com/minio/minio@latest, but there are no pre-compiled community binaries after 16 October 2025 and no maintenance. AIStor Free is free of charge and maintained, but it is commercially licensed, limited to a "Single-node deployment architecture", supported only through community Slack and documentation, and "Redistribution is prohibited."

Why did MinIO remove the admin console from the community edition?

The release notes of 24 May 2025 state that the "Embedded UI Console is now deprecated and moved to object-browser" and that "External IDP logins via LDAP/OIDC are removed as well; these are now available as part of the AiStor Product." The same note recommends that paying customers upgrade to AIStor. STS APIs were kept working so that anyone could "build UI in front."

What is the best MinIO alternative?

It depends on which half of MinIO you were using. For one node with an S3 API over a disk, look at versitygw or Garage. For many small files, SeaweedFS. For a cluster on Kubernetes alongside block and file, Ceph with Rook. For analytics over large files, Apache Ozone. For an S3 front end over other clouds, Zenko CloudServer. If you no longer want to run it at all, R2 and Tigris for free egress, B2 and Wasabi for cheap bulk, Hetzner and Scaleway for EU locations, and Out Plane if the bucket exists to serve an app you are already deploying.

What is the best open source MinIO alternative?

SeaweedFS (Apache-2.0) and Apache Ozone (Apache-2.0) are permissively licensed and clustered. Garage is AGPL-3.0, like MinIO, and is by far the smallest to operate. Ceph is dual-licensed under LGPL 2.1 or 3.0, with the Rook operator under Apache-2.0. Zenko CloudServer and versitygw are both Apache-2.0. If your reason for leaving was the AGPL rather than the archive, the Apache-2.0 projects are where to look.

Can I switch object storage without changing my application code?

Mostly. Every option here implements the S3 API, so you change the endpoint, the region, the key pair and sometimes the addressing mode. The things that usually need code changes are bucket-level features: lifecycle rules, object lock, server-side encryption settings, bucket policies and anything that used MinIO's admin or STS APIs. Out Plane's docs are a good illustration of what to check in a client: the region is signed as auto and addressing is virtual-hosted only, so path-style requests do not work.

Can I migrate from MinIO without downtime?

Close to it. Mirror the bulk while the old store still serves traffic, keep a mc mirror --watch or a repeating rclone sync running so the delta stays near zero, then take a short write pause for the final sync and the configuration change. Reads can continue from the old store throughout. The only unavoidable gap is the moment you flip the endpoint and redeploy.

How long does a migration take?

Transfer time, not tool time, dominates. Because rclone can only do server-side copies when both sides are the same remote, a cross-provider move streams every byte through the host running the copy. Measure one bucket first, then multiply. Running the copy on a server near one end, with 200 transfers and 200 checkers as the rclone docs suggest, is usually the difference between hours and days.

Does Out Plane replace a MinIO cluster?

It replaces the bucket, not the deployment model. Out Plane buckets are a managed S3 compatible service with per bucket access keys and unmetered egress, suited to applications that store and serve files. If you need an object store running on hardware you control, pick from the self-hosted section above, or deploy one on Out Plane from the MinIO template and keep owning its configuration and upgrades.

Sources

Facts checked on 15 November 2026 against the MinIO repository, MinIO RELEASE.2025-05-24, MinIO LICENSE, AIStor pricing, minio/mc, minio/minio-go, the mc mirror reference and mc alias set reference, the SeaweedFS repository and SeaweedFS S3 API, Garage, Garage design goals and the Garage mirror repository, the Ceph repository and Rook object storage docs, the Apache Ozone docs, Zenko, Zenko CloudServer, versitygw, Cloudflare R2 pricing, Backblaze B2 pricing, Hetzner Object Storage and its documentation, Scaleway storage pricing and Object Storage concepts, Tigris pricing, Wasabi pricing, the rclone S3 docs and rclone repository, the Out Plane bucket docs and the Out Plane pricing page. Licences come from each project's own repository. Prices and licences change, so confirm them before you commit.

Next Steps

Start from the two questions: does the data have to live on your hardware, and how much do you serve to the internet. Those answers cut this list of twelve down to two or three, and the rest is a copy job you can rehearse on one bucket.

If a managed bucket beside the app that writes to it is the shape you want, create one in the Out Plane console on the 14 day free trial, read the bucket docs for the auto region and virtual-hosted addressing, and check the pricing page. For wider comparisons, see the best PaaS providers, free PostgreSQL hosting and the alternatives hub.


Tags

minio
minio-alternatives
object-storage
s3
comparison
self-hosted
cloud-storage

Next steps


Related reading


Start deploying in minutes

Connect your GitHub repository and deploy your first application today. Start with a 14-day free trial.

Start 14-day free trial