llmfit
SHA-256A terminal tool that detects your hardware and right-sizes hundreds of LLM models to your RAM, CPU, and GPU. One command to find what runs on your machine. Supports multi-GPU, MoE architectures, and local providers. More practical than running benchmarks manually.
Smart Download
Download Download Version
v1.1.6 · 6.1 MB
Scans your hardware and tells you which LLMs will run well — no downloading, no guessing.
Core Features
- Auto-detects RAM, VRAM, CPU, and GPU backends — no manual configuration needed
- Scores hundreds of models on fit, speed, quality, and context
- Correctly handles MoE models (e.g., Mixtral, DeepSeek-V3) unlike alternatives that treat them as dense
- Interactive TUI (default) plus classic CLI mode for scripting and automation
- Built-in benchmark to measure real tok/s and contribute results back to the community
What It Can't Do
- •Speed estimates are based on a memory‑bandwidth model; real performance may vary due to system load. Use `llmfit bench` for actual numbers.
- •The model catalog is built-in; see custom-models docs to add models not in the database.
- •The TUI may require a modern terminal emulator (e.g., kitty) for full rendering.
Use Cases
- Quickly find the best LLM for your laptop or desktop without downloading models blindly
- Automate model selection in CI/CD pipelines or agent workflows using JSON output
Detailed Introduction
llmfit is a cross-platform terminal utility that solves the common pain point of figuring out which large language models can actually run on your local hardware. It scans your system (RAM, VRAM, CPU, GPU backends) and scores every model from its extensive catalog across four dimensions: memory fit, estimated speed, quality, and context length. Instead of guessing or downloading huge models only to find they won't load, you get an instant ranked list. It ships with a beautiful interactive TUI (default) and a CLI mode for scripting. Unique advantages: (1) it uses a memory-bandwidth model grounded in real community measurements, not just theoretical specs, so speed estimates are more accurate; (2) supports Mixture-of-Experts (MoE) models correctly, unlike alternatives like llm-checker which treat MoE as dense. llmfit also lets you benchmark real tok/s on your machine and share results back to the project via a PR, improving estimates for everyone.
Tags
Getting Started
Download installer
Click the button above to download the installer for your system
Install the software
Double-click the downloaded installer and follow the prompts
Install: `scoop install llmfit` (Windows), `brew install AlexsJones/llmfit/llmfit` (macOS/Linux), or curl script
Run: type `llmfit` to open the interactive TUI showing hardware info and ranked models
Advanced: use `llmfit bench` to measure real speed, or `llmfit recommend --json` for machine-readable output
- Install: `scoop install llmfit` (Windows), `brew install AlexsJones/llmfit/llmfit` (macOS/Linux), or curl script
- Run: type `llmfit` to open the interactive TUI showing hardware info and ranked models
- Advanced: use `llmfit bench` to measure real speed, or `llmfit recommend --json` for machine-readable output
SHA-256 checksum verified
Checksum extracted from GitHub official Release page
SHA256 Checksum
1e09232a128455596a2d348ab5893741d04b94aa6d924f1253462dc13304f7c6This checksum is extracted from the GitHub Release page. Verify file integrity after download.
All SHA-256 checksums on this platform are extracted from the project's official GitHub Release page, without any modification. You can independently verify them on the GitHub Releases page.
Open Source Transparency
View GitHub SourceUninstall Info
Uninstall depends on method: `scoop uninstall llmfit`, `brew uninstall llmfit`, or delete the binary and `~/.local/bin/llmfit` if installed manually.
No Extra Dependencies
Ready to use after download. No additional runtime required.
Similar Projects
opencode
OpenCode is a free, open-source AI coding agent that helps you write, refactor, and understand code. It offers two built-in modes: a full-access 'build' agent for active coding and a read-only 'plan' agent for safe exploration.
codex
OpenAI's lightweight coding agent that runs locally in your terminal, with no IDE dependency and seamless ChatGPT integration.
lazygit
A simple terminal UI for git commands