llama.jar
by Popr4x
llama.jar brings offline local language model inference into a Minecraft server, running GGUF models inside the game process.
What it does
Inside the Minecraft process, llama.cpp runs locally, letting players load and query open models like LLaMA-3, Qwen-2.5, Phi-3 and Mistral with no outside service involved. Inference works offline on the host's own hardware, and no API keys, subscriptions or outside network connections are involved.
Hardware acceleration is available for generation, with multi-core CPU execution alongside GPU offloading through CUDA, Metal and OpenCL/CLBlast, and configurations with more than one GPU are supported. Models are placed as GGUF files in the models folder inside the game directory, and the mod sets up its workspace on first launch.
The mod works on Fabric and Forge across 1.20.1, 1.21.1 and 1.21.11, and it is server-side only, so players join without installing anything. Operators manage loaded models and run prompts through in-game commands gated by permissions. Settings live in llamajar-common.toml and llamajar.json in the config directory, where performance parameters can be tuned. The mod is released under the GNU Affero General Public License v3.0 only and is not open source.
Key features
Offline model inference
Models run on the server's own hardware with llama.cpp, with no external network calls or subscriptions required.
GPU and CPU acceleration
Generation can use multi-core CPU execution together with GPU offloading, including support for multiple GPUs.
In-game prompt commands
Operators load models and send prompts to them from chat commands, with each action restricted by command permissions.
GGUF model loading
Any GGUF model file placed in the models folder can be loaded, and the performance parameters can be tuned in the generated config files.
Compatibility
Latest build for each game version and loader.
Only full game releases are shown. This mod publishes no snapshot builds.
Use cases
Who it’s for: Server owners who want local language models available through commands, and developers who want to experiment with GGUF models inside a Minecraft world. Players only join the server and need nothing installed.
Questions
Does a player need llama.jar installed to use it on a server?
Which loaders and versions does llama.jar support?
In the catalog
llama.jar is one of 3,514 Library mods on ModsFinder, and one of 2,348 Library mods for Fabric. 19,862 mods support Minecraft 1.21.11.
Similar mods
What the author says
llama.jar is a lightweight, high-performance Minecraft mod that brings the power of local Large Language Models (LLMs) directly into Minecraft using llama.cpp. Runs completely offline on your own hardware with no API keys, no subscriptions, and zero external network requests.
Official listing
- Modrinth page
- modrinth.com/mod/llama.jar
- Owner
- Popr4x
- Published
- 2026-05-19
- Updated on Modrinth
- 2026-05-19
- Facts verified
- 2026-10-15