App & Framework Hosting is now generally available in all 30+ regions.

Game Servers

Paper vs Spigot vs Forge vs Fabric: Which Minecraft Server Software?

By Adam Eastwood · · 8 min read

Quick answer

Choose Paper for a plugin server — it outperforms Spigot and runs the same plugins. Choose Fabric for lightweight mods and fast version updates, and Forge or NeoForge for large modpacks, since most big packs are built on them. Plain Spigot is rarely the right choice any more.

Every Minecraft server runs on one piece of server software, and the choice you make first — before plugins, before mods, before the world exists — decides what your server can become. Pick a plugin platform and your players join with an unmodified vanilla client; pick a mod loader and every player installs matching mods, in exchange for content that changes the game itself. This guide compares the five options that matter in 2026: Paper (and its fork Purpur), Spigot, Forge, NeoForge, and Fabric.

Whichever you land on, all of them run on standard Minecraft server hosting — the software choice changes your setup and RAM needs, not where you can host it.

What's the difference between plugins and mods?

This distinction drives everything else, so it is worth getting precise. Plugins run only on the server. They add commands, permissions, economies, minigames, and protection using the blocks and items that already exist in vanilla Minecraft. Because the client is untouched, anyone with a normal copy of the game can join. Mods change the game's code on both sides. They can add entirely new blocks, machines, dimensions, and mechanics — but every player must install the same mod loader and (mostly) the same mods, usually via a launcher profile or modpack.

Rule of thumb: if you want vanilla-with-extras that anyone can join, you want a plugin server. If you want Create contraptions, tech trees, or a modpack from CurseForge, you want a mod loader.

Is Paper better than Spigot?

For almost every server, yes. The history explains why: the original plugin platform was Bukkit; Spigot continued it with performance improvements and became the standard for years; Paper then forked Spigot and pushed much further — rewritten chunk loading, asynchronous operations where the vanilla server blocks, extensive exploit patches, and far more granular configuration. Paper remains compatible with the Bukkit and Spigot plugin API, so effectively the entire plugin ecosystem runs on it unchanged.

Spigot's remaining distinctives are mostly friction: you build the jar yourself with BuildTools, and you give up Paper's performance and configuration work. Paper also ships as a simple downloadable jar with an active release cadence. Unless a specific plugin documents a genuine incompatibility (rare, and usually a plugin relying on a vanilla bug Paper fixed), Paper is the default choice for a plugin server in 2026.

What about Purpur?

Purpur is a fork of Paper that keeps Paper's performance and plugin compatibility and layers on hundreds of optional gameplay toggles — ridable mobs, configurable mob AI behaviours, and countless small tweaks that vanilla hard-codes. Everything is off by default, so a fresh Purpur server behaves like Paper. It is popular with server owners who want fine-grained control without writing plugins. The trade-off is one more link in the update chain: Purpur updates after Paper, which updates after Mojang.

Forge vs Fabric vs NeoForge: which mod loader?

Forge is the veteran, with well over a decade of modding history behind it, and the majority of large, established modpacks were built on it. NeoForge forked from Forge in 2023, taking most of the active development community with it; new modpack development has largely moved there, and for recent Minecraft versions “Forge-style” packs are typically NeoForge packs. In practice you rarely choose between them in the abstract — the modpack you want to run tells you which loader and which exact version it needs.

Fabric took the opposite philosophy: a tiny, lightweight loader plus a separate API mod, designed for fast updates and minimal overhead. Fabric mods typically appear within days of a new Minecraft release, and the ecosystem is strong on performance and quality-of-life — server-side optimisation mods like Lithium, and client-side renderers like Sodium, come from this world. Big kitchen-sink packs are less common on Fabric than on Forge/NeoForge, though the gap has narrowed.

How do the five options compare?

Minecraft server software compared, as of August 2026.
SoftwareTypeClient needsStrengthsWeaknesses
PaperPlugin serverVanilla clientBest plugin-server performance, huge plugin ecosystem, active developmentNo mods; a handful of vanilla-mechanic edge cases differ
PurpurPlugin server (Paper fork)Vanilla clientEverything Paper has, plus extensive optional gameplay togglesUpdates lag slightly behind Paper
SpigotPlugin serverVanilla clientMature, defined the plugin API standardSuperseded by Paper on performance and features; BuildTools setup
Forge / NeoForgeMod loaderMatching mods installedRuns the deepest catalogue of large modpacks; NeoForge carries active development forwardHeavier, slower to update after new Minecraft versions, higher RAM needs
FabricMod loaderMatching mods installedLightweight, updates fast, excellent performance-mod ecosystemFewer big kitchen-sink packs than Forge/NeoForge

Can you mix plugins and mods?

Mostly, no — and you should be sceptical of anything promising otherwise. Hybrid servers (projects like Mohist and similar) attempt to run Bukkit-style plugins on a Forge base, but they sit on a fragile intersection of two APIs that were never designed to coexist. The practical result is subtle breakage: plugin protection that doesn't see modded blocks, world corruption edge cases, and no upstream support when things go wrong. Two sturdier patterns exist. On Fabric, some server owners run server-side-only mods that replicate common plugin functions. And on any loader, several mod-loader-native permission and protection tools now cover the basics that plugins traditionally handled. If you genuinely need both ecosystems at full strength, run two servers behind a Velocity proxy rather than one hybrid.

Which should you pick?

  • Survival server for friends, vanilla feel: Paper. You get vanilla gameplay, better performance than the official jar, and the option of plugins later.
  • Public community server (ranks, economy, claims, minigames): Paper, or Purpur if you want its extra toggles.
  • A specific modpack: whatever loader the pack was built for — the pack page tells you. Most large packs mean Forge or NeoForge; check the RAM recommendation and see our modded RAM guide for sizing.
  • Custom light modding or maximum performance: Fabric with a curated mod list.
  • Plain Spigot: only if something specific forces it — otherwise Paper does everything Spigot does, faster.

Does the software change performance and RAM needs?

Noticeably. Paper and Purpur are meaningfully lighter than the vanilla jar under load — their chunk handling and entity tuning are the main reason a busy plugin server holds 20 TPS where vanilla would stutter. On the modded side the loader itself matters less than what it carries: a Forge or NeoForge kitchen-sink pack needs 8GB or more simply because hundreds of mods load their content into memory, while a lean Fabric server with a handful of optimisation mods can run comfortably in 3–4GB. Two practical notes: server-side performance mods (Lithium and friends) go on the server, but Sodium and other renderers are client-only — installing them server-side does nothing — and whatever you run, size the allocation using real player count and pack weight rather than the loader's name.

Switching later: what carries over?

Within the plugin family, migration is painless: Spigot → Paper → Purpur all read the same world format and the same plugins folder, so you can upgrade in place (take a backup first, always). Between a plugin server and a mod loader, the world's vanilla terrain survives but everything added by the other side does not — modded blocks vanish from a world opened without their mods, and plugin data stays behind. Treat plugin-vs-mods as the one decision to get right before your community builds anything they love.

The good news: none of this is irreversible at the start. Spin up a server, try Paper for a weekend, try a small Fabric pack the next — worlds are cheap until they are full of memories.