Skip to content
Get packs

Guide · Open source

How to use llmfit: find which local LLMs your hardware can run

Updated · written with AI from the project's own documentation and checked against our catalogue data

llmfit is an open-source, MIT-licensed command-line tool written in Rust. It checks your CPU, system RAM, GPU and VRAM, then recommends open-source large language models that should run comfortably on that hardware. This guide shows you how to install it, run its interactive and plain command-line modes, and read the results so you avoid downloading models your machine cannot handle.

llmfitAlexsJones/llmfit · ★ 38.6k · MIT

Quick start

  1. 1Install llmfit. Run the WinGet, Homebrew or uv command for your system.
  2. 2Launch the TUI. Run llmfit with no flags to see your detected hardware and scored models.
  3. 3Filter the list. Press / and type a model name, family or quantization to narrow the results.
  4. 4Get plain or JSON output. Run llmfit recommend, or add --json, to print recommendations for scripts or review.
  5. 5Check disk space if needed. Run llmfit storage with --keep to estimate SSD capacity for several models.

Install llmfit with your package manager

Pick the command that matches your operating system. The README lists these install methods; use the one you already have available.

  • Windows (WinGet): run the WinGet command below.
  • macOS or Linux (Homebrew): run the Homebrew command below.
  • Any platform with uv: run the uv command below. Use uvx llmfit to run it without installing.
winget install --exact --id AlexsJones.llmfit brew install AlexsJones/llmfit/llmfit uv tool install -U llmfit

Start the interactive terminal interface

Run llmfit with no flags to open the interactive TUI. It shows your detected specs at the top and every model scored for fit, speed, quality and context. Use the arrow keys or k and j to move through the list, and / to filter models by name, family or quantization.

  • h: show help and keybindings
  • b: open community benchmarks
  • I: open live inference benchmarks
  • Esc: clear search or go back
llmfit

Get recommendations from the command line

If you prefer plain output or want to script the results, use the recommend command. Add --json to get the system profile and recommendations as raw JSON, which you can pipe into other tools.

  • llmfit recommend prints hardware telemetry and recommended models to standard output.
  • llmfit recommend --json outputs the same information as JSON.
llmfit recommend llmfit recommend --json

Estimate storage and run the HTTP API

The storage command estimates SSD capacity for keeping a set of runnable models, which helps teams sizing disks. The serve command starts a native HTTP API server; the README documents the /api/v1/system and /api/v1/models endpoints.

  • Storage estimate for three models, largest first, as JSON.
  • Serve the API on port 8787 with the host set to 0.0.0.0.
llmfit storage --keep 3 --selection largest --json llmfit serve --host 0.0.0.0 --port 8787

Run llmfit in Docker

The project publishes a multi-architecture image at ghcr.io/alexsjones/llmfit. Without flags it prints JSON from llmfit recommend. Pass the global --tui flag to launch the interactive interface inside the container. To run the web UI and API, publish port 8787 and use the serve command.

docker run -it --rm ghcr.io/alexsjones/llmfit --tui docker run -d -p 8787:8787 ghcr.io/alexsjones/llmfit serve

Build from source

If you want to build the binary yourself, clone the repository and build it with Cargo. The README says the binary is written to target/release/llmfit.

git clone https://github.com/AlexsJones/llmfit.git cd llmfit cargo build --release

What people use it for

  • Deciding which local LLM to download for a laptop or desktop before spending time on a large download.
  • Comparing quantized models such as GGUF, AWQ, GPTQ and EXL2 for one specific machine.
  • Sizing hardware or SSD space for a team that plans to keep several local models.
  • Feeding recommendation data into dashboards or deployment pipelines through the JSON output or REST API.

Frequently asked questions

Is llmfit free?

llmfit is released under the MIT licence, as stated in the project's README and listing. The MIT licence permits use, modification and distribution, so check the LICENSE file in the repository for the full terms.

How do I install llmfit on Windows?

The README gives the WinGet command winget install --exact --id AlexsJones.llmfit. It also lists Scoop as an option with scoop install llmfit.

How do I install llmfit on macOS or Linux?

Use brew install AlexsJones/llmfit/llmfit for the prebuilt binary through Homebrew. The README also lists a MacPorts command and a quick install script, and offers release binaries on GitHub Releases.

What hardware does llmfit check?

It detects CPU cores, system RAM, discrete and integrated GPUs, VRAM and unified memory architecture, including NVIDIA CUDA, Apple Silicon, AMD ROCm and Intel OneAPI setups.

Can llmfit tell me exact speeds?

Its speed figures are estimates projected from model and hardware data. The README says you can benchmark a model on your own machine, and those measured results can replace estimates in the fit table.

Why is my Windows binary unsigned?

The README says Windows binaries are signed only when the release's sign-windows job completes successfully. A release may publish an unsigned Windows artifact if signing is skipped or fails, so verify the signature yourself if you need one.

New agent skills, repos and AI tools — every week

The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.

Our verdict

llmfit suits people choosing a local LLM for their own laptop or desktop, developers comparing quantized models, and teams sizing hardware. If you only use hosted APIs and never run models locally, or you need exact speed numbers without benchmarking, pick a different tool or plan to measure results yourself.

Install options & downloads