# llmfit
llmfit is an [[Open Source]] [[Command Line Interface (CLI)]] utility written in [[Rust]] that matches [[Large Language Models (LLMs)]] to your hardware. It detects your system's RAM, CPU, and GPU resources, then scores and ranks hundreds of models to recommend which ones will actually run well on your machine.
## Key features
- **Multiple interfaces**: interactive [[Terminal User Interface (TUI)]] with vim-like navigation, classic CLI with table output, REST API, and web dashboard (port 8787)
- **Hardware detection**: multi-GPU support (NVIDIA, AMD, Intel Arc, Apple Silicon, Ascend), automatic backend identification (CUDA, Metal, ROCm)
- **Dynamic quantization selection**: automatically picks the best quality quantization that fits your available memory
- **Mixture-of-Experts support**: handles MoE architectures with expert offloading calculations
- **Multi-dimensional scoring**: Quality, Speed, Fit, and Context window dimensions weighted by use case
- **Runtime integration**: detects installed models from [[Ollama]], llama.cpp, MLX, Docker Model Runner, and [[LM Studio]]
- **Plan mode**: inverse analysis that estimates required hardware for a specific model configuration
## How it works
1. **Hardware detection** . probes system resources and GPU capabilities
2. **Database lookup** . loads embedded model metadata from [[HuggingFace]] (600+ models)
3. **Memory calculation** . computes requirements across quantization hierarchy (Q8_0 to Q2_K)
4. **Speed estimation** . uses GPU bandwidth tables: `(bandwidth_GB/s / model_size_GB) x efficiency_factor`
5. **Scoring and ranking** . applies weighted dimensions by use case, sorts by composite score
Fit modes classify models as GPU-native, MoE-offload, CPU+GPU hybrid, or CPU-only with levels from "Perfect" to "Too Tight".
## Installation
Available via Homebrew (`brew install llmfit`), Scoop, Docker, curl install script, or build from source with [[Cargo (Rust)]].
## References
- https://github.com/AlexsJones/llmfit
## Related
- [[Large Language Models (LLMs)]]
- [[Ollama]]
- [[LM Studio]]
- [[HuggingFace]]
- [[Rust]]
- [[Terminal User Interface (TUI)]]
- [[Command Line Interface (CLI)]]
- [[Model routing]]
- [[Open Source]]