PumpkinMC Beta Is Live After 28 Months of Rust Development

PumpkinMC has entered public beta after 28 months of development, with version 1.0 due on September 4, 2026 at 8:00 PM CEST. Its name creates an obvious conflict of interest for MrPumpkin, but I put the ceremonial gourd aside while reading the docs and release announcement, because this is a serious attempt to rebuild the Minecraft server from scratch in Rust.

What PumpkinMC is

Paper, Spigot and most familiar server forks build on Minecraft's Java server. PumpkinMC reimplements the server, giving its developers control over networking and game logic from the start rather than working around an existing main-thread design.

The administration docs describe multithreaded execution across modern CPU cores and low memory use. Rust produces a native executable too, so there is no Java runtime, heap size or garbage collector to configure. Admins still get familiar basics: pumpkin.toml configuration, Velocity and BungeeCord forwarding, and downloads for Windows, Linux and macOS.

Plugins without the JVM

Existing Bukkit and Paper .jar plugins do not run directly in Pumpkin, which is the first migration problem for established servers. Pumpkin instead loads sandboxed WebAssembly plugins, and its developer guide covers Rust, Python, C#, C, C++, Go and Kotlin. Each route produces a .wasm file, giving developers seven language choices without tying the plugin system to the JVM.

The API borrows ideas from Bukkit and Spigot, although Pumpkin's documentation warns that it is still early and may change. Its plugin catalog is also far smaller than Bukkit's, so admins should check for replacements before moving a live server, especially if that server has somehow developed a load-bearing fishing economy.

Performance and the benchmark problem

Pumpkin's home page advertises startup around 5 milliseconds, roughly 100 MB of idle memory and about 0.1 percent idle CPU use. Those are Pumpkin's own figures, and the published benchmark page says its comparison with Vanilla, Paper, Purpur and Minestom is currently unfair because Pumpkin implements fewer features. The test also used stationary players and only eight loaded chunks, while the numeric result cells are blank at the time of writing, so the headline numbers should not be used to size a busy production server yet.

Beta, 1.0 and vanilla parity

Pumpkin's target for 1.0 is a game that can be beaten from beginning to end. Perfect vanilla world generation is not part of that target, although the project says one-to-one generation and deeper parity remain planned. Recreating years of mob AI, redstone, combat and edge cases takes time, so players should expect the beta to expose behavior that still feels wrong.

The team wants reports of crashes, bottlenecks and odd behavior through GitHub Issues, preferably with logs and reproduction steps. PumpkinMC is worth trying on a test instance or new world, but I would not move a long-running Paper network until its required plugins and important gameplay have been tested.

The beta can be downloaded now, and I am mostly watching the plugin market from here because server software becomes much easier to judge once developers start feeding it their unreasonable ideas and finding the gaps.

Keep your pumpkin lit and your diamonds close. - MrPumpkin