AioLMView on GitHub
Changing language opens this page in the selected language.

Local models. One workspace.

Discover GGUF models, manage llama.cpp runtimes, chat locally, and measure performance with AioLM.

View on GitHubExplore benchmarks
> powershell -ExecutionPolicy Bypass -Command "irm https://github.com/aiolm/AioLM/releases/latest/download/install.ps1 | iex"
The AioLM desktop app on its Run a model page, listing four public GGUF demonstration models with Qwen3-8B-Instruct loaded on a llama.cpp runtime.
Run a model in the AioLM desktop app: your local GGUF library, the loaded model and its llama.cpp runtime in one window. Shown with public demonstration models.

From model to measurement

A simple workflow for local LLMs, designed for builders.

  1. Choose your model

    Discover GGUF models from the community and add them to your library.

  2. Run it your way

    Manage llama.cpp runtimes and configure settings that fit your hardware and workflow.

  3. Measure and share

    Run benchmarks, review results with full context, and share your configuration with others.

Frequently asked questions

What is AioLM?

AioLM is short for All-in-One LM: a desktop workspace for local language models. It brings GGUF model discovery, llama.cpp runtime management, local chat and performance benchmarks into one application.

Which operating systems does AioLM support?

Windows x64, macOS 13.3 or newer (Apple silicon and Intel) and Linux x86_64 (Ubuntu 24.04 or newer) builds are available. Apple silicon Macs can use Metal GPU acceleration; Intel Macs run models on the CPU. Each platform's installation guide covers the release assets.

What do the public benchmarks measure?

Public benchmarks are self-reported measurements shared by AioLM users. Each result includes its hardware, runtime, measurement method and workload. Compare those conditions before comparing performance; these results are not an independently verified ranking.

Do I need an account to browse benchmarks?

No account is required to browse public benchmark results. Publishing uses a verification step, and published results are managed with a recovery code.

Understand the setup behind the numbers

Published benchmarks include the measurement method, workload and hardware so configurations can be compared.

Browse benchmarks