Built on Phi-3.5-mini-instruct

Chat with an LLM Trained on All OSRS Content

A highly-performant, lightweight generative AI model trained exclusively on Old School Runescape knowledge. Perfect for OSRS enthusiasts who want accurate, context-aware conversations.

About OSRS LLM

Built specifically for the OSRS community with comprehensive game knowledge from various sources.

Designed to run on basic consumer hardware including lower VRAM GPU's, OSRS LLM delivers expert answers about quests, items, skills, and game mechanics without relying on cloud services.

OSRS Expert Knowledge

Trained on all OSRS content including quests, items, skills, mechanics, and community knowledge up to date.

Lightweight & Fast

Built on Phi-3.5-mini-instruct, runs efficiently on consumer hardware including 12GB VRAM GPUs.

Run Locally

Host it yourself with complete privacy. Works with OpenWebUI, Ollama, and all standard LLM chat applications.

How It Works

Get up and running in minutes with your own OSRS AI assistant

1

Download Model

Get the model files from HuggingFace or clone the GitHub repository

2

Choose Your App

Pick OpenWebUI, Ollama, or any compatible LLM chat interface

3

Load the Model

Import the OSRS LLM into your chosen chat application

4

Start Chatting

Ask anything about OSRS and get expert, contextual answers

Built on Phi-3.5-mini-instruct

Microsoft's Phi-3.5-mini is a lightweight open model built upon high-quality datasets including synthetic data and filtered publicly available websites. It excels at reasoning and instruction adherence with robust safety measures.

  • 128K token context length for long conversations
  • Runs on consumer hardware (12GB VRAM)
  • Enhanced with supervised fine-tuning and DPO
  • Customizable training pipeline for more context

Compatible Platforms

🌐

OpenWebUI

Full-featured web interface

🦙

Ollama

Simple CLI and API

🤗

HuggingFace

Direct model access

Ready to Chat with OSRS AI?

Get the model from GitHub and start running your own OSRS expert locally. Perfect for answering questions about quests, items, skills, and everything Gielinor.

Additionally, reference the training pipeline to add any additional context or inference desired to train your local version of the model. Runs with full privacy on your own hardware.