Open SourceLocal8k

KoboldCpp

A single-file, llama.cpp-based runner with a built-in UI and image generation.

KoboldCpp is an easy-to-deploy, single-executable inference runner built on llama.cpp that adds a built-in web UI, persistent chat, and support for image generation and speech models. It runs GGUF models across CPU and many GPU backends with no installation, making it popular for local chat, storytelling, and roleplay use cases. It also exposes a KoboldAI and OpenAI-compatible API for integration.

View repo

Repository

LostRuins/koboldcpp

Language

C++

What you'd build with it

Running local GGUF models from a single portable binary with zero install
Self-hosting a chat or storytelling UI with persistent context on modest hardware
Combining text, image, and speech generation behind one local API endpoint

Repository

Language

What you'd build with it

Tags