If you're spinning up a Minecraft server for the first time, the jar you pick decides a lot about how the next few months go. Plugins or mods? Better performance or more freedom? This is a plain rundown of vanilla, CraftBukkit, Spigot, Paper, Fabric and Forge so you can pick the one that actually fits what you want to build.
Start with the two big families
Before the names, here's the split that matters. There are two separate worlds in Minecraft server software, and they don't mix.
One side is built for plugins. That's CraftBukkit, Spigot and Paper. The other side is built for mods. That's Forge and Fabric. Vanilla sits on its own, running the game exactly as Mojang ships it.
Plugins change what the server does. Mods can change what the game is, on both the server and the client. We'll get into why that difference is the whole ballgame, but keep it in your head as we go through each option.
Vanilla
Vanilla is the official server jar straight from Mojang. No extras, no surprises. You download it, accept the EULA, set a few things in server.properties, and you've got the exact game everyone gets in singleplayer, just shared.
The upside is honesty. Updates land the same day Mojang releases them. Nothing breaks because of a third party. If you want a small survival world for you and a handful of friends, vanilla is genuinely fine and there's no shame in it.
The downside shows up fast once you grow. There's no way to add a permissions system, no economy, no land claims, no anti grief tools. You can't even set a spawn protection rule beyond the one basic radius. And performance is, honestly, the weakest of the bunch. The default server isn't tuned for 30 people fighting in one chunk.
Use vanilla when you want the pure game and a tiny group. Move on the moment you find yourself wishing you could ban someone with a command instead of editing a file by hand.
CraftBukkit and Spigot
Years ago, the community wanted a way to add features without rewriting the game. Bukkit was the answer. It's an API, a set of hooks that let small programs (plugins) talk to the server. CraftBukkit is the server that implements that API.
CraftBukkit on its own is mostly a piece of history now. Almost nobody runs it directly. What people mean when they say "Bukkit server" is usually Spigot, which is CraftBukkit with a pile of performance patches and extra config options bolted on. Spigot gave you spigot.yml, where you could tune things like entity activation ranges and view distance to claw back some lag.
Spigot was the default choice for a long time. It still runs basically every plugin out there. But here's the thing: there's almost no reason to pick Spigot in 2026 over Paper, which is built on top of it and is better in nearly every way. So let's talk about Paper.
Paper
Paper is a fork of Spigot. It keeps full plugin compatibility (a plugin written for Spigot runs on Paper without changes) but rewrites huge chunks of the internals for speed and adds fixes Spigot never bothered with.
In practice Paper is the one we recommend to most people running a plugin server, and it's what a lot of hosts default to. You get extra config in paper-global.yml and paper-world-defaults.yml (older versions used a single paper.yml), better chunk loading, and far fewer of the weird dupe bugs and lag spikes that plague vanilla and plain Spigot.
If you want the popular plugins, this is your jar. We're talking:
- LuckPerms for ranks and permissions
- EssentialsX for homes, warps, kits and basic economy
- WorldEdit and WorldGuard for building and region protection
- Vault as the glue between economy and permission plugins
- CoreProtect for rolling back grief
The catch with the whole Bukkit and Spigot and Paper line is that plugins can't add genuinely new content the player downloads. No new blocks, no new mobs with custom models, no new dimensions in the way a mod does it. Plugins work within the existing game. They're incredibly flexible inside those walls, but the walls are there.
A quick warning on Paper updates: it sometimes removes a vanilla bug that a redstone contraption secretly relied on, or changes a default that affects gameplay. Read the release notes before you bump a major version on a server people care about.
Forge
Now the other family. Forge is the old guard of modding. If you've ever installed a big modpack from CurseForge, it was probably running on Forge. It lets mods change the game deeply, and crucially, the same mods usually have to be installed on the client too.
That's the core difference between a plugin and a mod. A player joining your Paper server needs nothing but vanilla Minecraft. A player joining your Forge server with a 150 mod pack needs that exact pack installed locally, or they can't connect. The server and client have to agree on what reality looks like.
Forge suits modpacks. Think tech and magic packs, Pixelmon, RLCraft, the big kitchen sink experiences. The trade off is weight. Forge plus a heavy pack eats RAM. A pack that vanilla would run at 2 GB might want 6 GB or 8 GB, and it's more sensitive to a single badly behaved mod tanking your tick rate.
Fabric
Fabric is the newer, lighter modding option. It's modular, updates to new Minecraft versions quickly, and the performance focused crowd loves it. A Fabric server running mods like Lithium, Starlight and FerriteCore can run noticeably leaner than the same idea on Forge.
Fabric needs a small loader mod called the Fabric API for most mods to work, and the mod ecosystem, while excellent, is smaller than Forge's for the giant established packs. If you want a slim performance setup, a smaller curated mod list, or a vanilla plus quality of life server, Fabric is a great shout. For the huge classic modpacks, Forge still has more.
One more name you might hear: NeoForge, a community fork of Forge that's taken over a lot of newer pack development. If a modern pack tells you it needs NeoForge instead of Forge, that's why. It behaves much like Forge for our purposes here.
So can you mix plugins and mods?
Mostly, no, and this trips people up. A Paper plugin won't run on Fabric. A Forge mod won't run on Paper. They're different engines.
There are hybrid servers that try to bridge the two, like Arclight or Mohist, which run Forge mods and Bukkit plugins together. They exist and some people swear by them. In our experience they're fiddly, they lag behind on updates, and they can produce bugs that neither the mod authors nor the plugin authors will help you with. If you can avoid a hybrid, do. Pick a side.
Performance in the real world
People ask which one is fastest like it's a single number. It isn't, but here are honest rules of thumb.
For a plain plugin server, Paper beats Spigot beats vanilla, clearly. Paper's chunk and entity handling is the difference between 20 players being smooth and being a slideshow on the same hardware.
For modded, a tuned Fabric setup with performance mods is usually the lightest. Forge and NeoForge are heavier by nature, and a big pack will always cost more RAM and CPU than any plugin server doing similar player counts.
Whatever you run, install spark. It's a profiler that works across Paper, Fabric and Forge, and it'll tell you exactly what's burning your tick time instead of you guessing. Type /spark profiler when things feel bad and read the report. It has saved us more debugging hours than anything else.
Rough starting points for RAM, assuming decent NVMe backed hosting:
| Setup | Players | Sensible RAM |
|---|---|---|
| Vanilla or Paper, light plugins | up to 10 | 2 GB to 4 GB |
| Paper, full plugin suite | 20 to 40 | 4 GB to 8 GB |
| Fabric, performance mods | 10 to 20 | 4 GB to 6 GB |
| Forge or NeoForge, big pack | 5 to 15 | 6 GB to 12 GB |
Don't just throw 16 GB at a small server, by the way. Past what you need, extra heap can actually make garbage collection pauses worse. More isn't always better.
Switching later and keeping your world
Good news here. Moving within the plugin family is easy. Going from vanilla to CraftBukkit to Spigot to Paper keeps your world fine, because they all read the same world format. You just stop the server, swap the jar, and start it again. Going forward through that line is painless.
Going from Paper back to vanilla also works, since the world data is standard. Just remember any plugin specific data (a plugin economy, region files from WorldGuard) won't mean anything to vanilla.
The harder jumps cross the family line. Moving a plugin world to a modded server, or the reverse, can work for the base terrain but anything the other side added gets weird or lost. And the big one: you cannot cleanly remove a heavy modpack from a world that already used its content. If players built with modded blocks and ores, taking the mods away leaves holes and errors where that content used to be.
So the practical advice is simple. Always back up the whole server folder before any switch. The world, world_nether and world_the_end folders are the bits you can't replace. Copy them somewhere safe, then experiment. A control panel that has a one click backup button (ours does, and so do most decent hosts) makes this a non event.
A simple way to decide
Strip it all back and it's not a hard call:
- Tiny survival with friends, pure game? Vanilla.
- A real server with ranks, claims, an economy, anti grief, a public IP? Paper. This is most people.
- You want to play a big modpack with custom content? Forge or NeoForge, whatever the pack asks for.
- You want a lean modded or vanilla plus setup with great performance? Fabric.
Most folks reading this want Paper and just didn't know the word yet. Start there, add LuckPerms and EssentialsX, and grow from there. You can always back up and try something heavier once you know what your players actually want. Pick the jar that matches the server in your head, keep a backup before you change anything, and you'll be fine.
