Guidefor server owners4 min read
Which Minecraft server software should you run in 2026?
Vanilla, Paper, Purpur, Folia, Fabric, NeoForge or a proxy. What each one is actually for, what it costs you, and the one question that decides between them.
The one question that decides it
Before comparing anything, answer this:
Do the people joining your server have to install something?
If the answer is no, you want a plugin server: Paper, or a fork of it. Players join with an ordinary Minecraft client and never think about it.
If the answer is yes, you want a mod loader: Fabric or NeoForge. Everyone installs the same mod set, and in exchange you can change the game itself.
Almost every other decision follows from that one, and getting it backwards is the most expensive mistake on this page, because moving between the two later means rebuilding your whole feature set.
The plugin side
Vanilla
The server Mojang publishes. No plugins, no API, no configuration beyond
server.properties.
Run it when you want exactly the game as shipped and nothing else, or when you are testing whether a bug is yours or Mojang's. Do not run it for a public server: it has no permission system, no protection against grief, and it is slower than Paper at the same player count.
Paper
The default answer for a plugin server, and it has been for years.
Paper is a fork of Spigot, which is a fork of Bukkit. It gives you a large plugin API, a long list of performance fixes, patched exploits, and configuration for behaviour vanilla hard-codes. Most plugins target Paper first.
What it costs you: a handful of vanilla behaviours differ by default. If you have technical players building redstone machines that depend on exact vanilla tick behaviour, they will notice, and some of it is configurable back.
Build history and checksums are on the Paper page.
Purpur
Paper plus several hundred more configuration options. Ridable mobs, tweakable mechanics, toggles for things Paper leaves alone.
Run it when you want the knobs. It tracks Paper closely, so plugin compatibility is effectively identical. The trade is that a new Minecraft version reaches Paper before it reaches Purpur, because Purpur has to rebase.
Folia
Paper, but with the world split into regions that tick in parallel across CPU cores.
This is not a general performance upgrade and it is important to understand why. Folia helps when your players are spread out, because separate regions can tick at the same time. It does nothing when everyone is standing at spawn, because that is one region on one thread. It also breaks plugins that assume a single main thread, which is most plugins ever written.
Run it if you have hundreds of players spread across a large map and you have checked that your plugins are Folia-aware. Otherwise run Paper.
Spigot and Bukkit
Still alive, still work, and Paper is a superset of both. There is no reason to start a new server on either in 2026. If you are already on one, moving to Paper is a jar swap.
The mod side
Fabric
Lightweight, fast to update, and usually the first loader with a build for a new Minecraft version. The ecosystem leans towards performance mods, client-side improvements and smaller focused mods.
Fabric needs the Fabric API mod installed alongside it for most mods to work, which catches people out on their first install.
NeoForge
The successor to Forge, and where the large content mods live. If you want a hundred-mod kitchen-sink pack with machines, magic and dimensions, this is usually where those mods are.
Heavier than Fabric and slower to reach a new Minecraft version, because there is far more API surface to port.
Quilt
A fork of Fabric with a different governance model and some API differences. Most Fabric mods run on it. Smaller ecosystem. Run it if you specifically want it.
Current support for each loader against the newest Minecraft release is on the loaders page.
Proxies
A proxy is not server software. It sits in front of several servers and moves players between them without disconnecting.
- Velocity is the current answer. Modern, fast, actively maintained.
- BungeeCord is the old answer and still widely deployed.
- Waterfall was a BungeeCord fork and has been superseded by Velocity.
You need one when you have more than one server and want players to hop between them: a lobby, a survival world, a minigame server. You do not need one for a single server, where it only adds a hop and another moving part. See setting up a proxy network.
What about running both mods and plugins?
Hybrid servers exist. They are the least reliable choice here, because a hybrid has to track two independently moving APIs and stay compatible with both, which means it is usually the last thing to support a new Minecraft version, sometimes by many months.
If you genuinely need both, know that you are trading version currency for it. If you think you need both, check first whether a mod does what your plugin does, or the other way round. Usually one exists.
A short decision list
| If this is true | Run this |
|---|---|
| Public server, players install nothing | Paper |
| As above, and you want more knobs | Purpur |
| Hundreds of players spread over a big map | Folia |
| Small friends server with mods | Fabric |
| Big content modpack | NeoForge |
| Several servers, one player base | Velocity in front of them |
| Testing whether a bug is vanilla | Vanilla |
Whatever you pick, check it has a stable build for the Minecraft version you want before committing. Compatibility lists every project against every recent release, with the evidence shown for each claim.