{
    "products": [
        {
            "slug": "koboldcpp",
            "name": "KoboldCpp",
            "kind": "app",
            "website": "https://github.com/LostRuins/koboldcpp",
            "page": "https://listme.name/software/koboldcpp",
            "tagline": "Single-file GGUF runner with the KoboldAI Lite web interface",
            "summary": "KoboldCpp is an AGPL-3.0 program that runs GGUF models locally from a single file, with a bundled KoboldAI Lite web interface. It runs on Windows, Linux and macOS and also generates images, speech and music with compatible models.",
            "developer": "LostRuins",
            "platforms": [
                "windows",
                "linux",
                "macos"
            ],
            "pricing": "open-source",
            "price_note": "AGPL-3.0 software is free to run locally",
            "free_to_use": true,
            "open_source": true,
            "license": "AGPL-3.0",
            "source": "https://github.com/LostRuins/koboldcpp",
            "strengths": [
                "One file with no install step",
                "Runs on CPU or GPU",
                "AGPL-3.0 open source"
            ],
            "limits": [
                "AGPL-3.0 has strong copyleft terms",
                "Image, video and music need separate compatible models"
            ],
            "checked": "2026-10-10",
            "connected_to_publisher": false,
            "overview": "KoboldCpp runs GGUF language models on CPU or GPU and includes the KoboldAI Lite web interface for chat and story writing. The project describes a single file with zero install, and it ships ready-to-run downloads for Windows, Linux and macOS. The README lists text, image, video, speech and music generation with compatible models.\n\nThe README also lists an integrated agent, image generation and editing with models such as SDXL and Flux, and many additional features. The repository is AGPL-3.0 licensed and had about 12,000 stars when checked. Users who want a command-line server may compare llama.cpp or Ollama, which are both in this ranking."
        },
        {
            "slug": "llama-cpp",
            "name": "llama.cpp",
            "kind": "app",
            "website": "https://github.com/ggml-org/llama.cpp",
            "page": "https://listme.name/software/llama-cpp",
            "tagline": "C and C++ engine for running local LLMs, with a server and tools",
            "summary": "llama.cpp is an open-source C and C++ engine for running large language models locally, with a server, a library API and command-line tools. It is released under the MIT license.",
            "platforms": [
                "cli"
            ],
            "pricing": "open-source",
            "price_note": "Free and open source under the MIT license",
            "free_to_use": true,
            "open_source": true,
            "license": "MIT",
            "source": "https://github.com/ggml-org/llama.cpp",
            "strengths": [
                "MIT license with source on GitHub",
                "Runs many GGUF models on local hardware",
                "Includes a server and a library API"
            ],
            "limits": [
                "Needs command-line and build knowledge",
                "No built-in graphical app"
            ],
            "checked": "2026-10-10",
            "connected_to_publisher": false,
            "overview": "llama.cpp is an inference engine for large language models, written in C and C++ and maintained under the ggml organization. Its repository describes it as LLM inference in C and C++ and includes a server, a library API and command-line tools for running GGUF model files on local hardware.\n\nMany other local AI tools build on it. The Jan and GPT4All projects in this category and the llamafile project all list llama.cpp among their foundations. Because it is a low-level engine, llama.cpp suits developers and people who are comfortable with the command line. Users who want a graphical app are better served by Jan or GPT4All. The MIT license allows commercial use and modification."
        }
    ],
    "differences": [
        "Only KoboldCpp runs on Windows.",
        "Only KoboldCpp runs on macOS.",
        "Only KoboldCpp runs on Linux."
    ],
    "shared_rankings": [
        {
            "list": "local-llm-runners",
            "name": "Local LLM runners",
            "page": "https://listme.name/best/local-llm-runners",
            "of": 10,
            "results": [
                {
                    "product": "koboldcpp",
                    "rank": 6,
                    "score": 84.5,
                    "verdict": "Single-file AGPL-3.0 runner for GGUF models, with the KoboldAI Lite interface and image, speech and music generation on Windows, Linux and macOS. Its copyleft license and wider feature set place it below Ollama.",
                    "best_for": "Single-file local writing and chat with GGUF models",
                    "spec": {
                        "interface": "Web interface (KoboldAI Lite)",
                        "model_format": "GGUF",
                        "cloud_option": "Not listed"
                    }
                },
                {
                    "product": "llama-cpp",
                    "rank": 2,
                    "score": 90,
                    "verdict": "The engine many local tools build on, with GGUF support, a server and a library API. It needs command-line comfort, so beginners will find Ollama or Jan easier.",
                    "best_for": "Developers who want a local inference engine",
                    "spec": {
                        "interface": "Command line and server",
                        "model_format": "GGUF model files",
                        "cloud_option": "Not applicable"
                    }
                }
            ]
        }
    ],
    "ranks_higher_in_closest_shared_list": "llama-cpp",
    "page": "https://listme.name/compare/koboldcpp-vs-llama-cpp",
    "json": "https://listme.name/api/v1/compare/koboldcpp-vs-llama-cpp.json",
    "notes": [
        "Scores are editorial and only comparable inside one list. Products with connected_to_publisher true receive a disclosed ranking lift: https://listme.name/methodology"
    ],
    "source": "https://listme.name",
    "license": "Free to read and cite with attribution to listme.name.",
    "docs": "https://listme.name/developers"
}