FitLLM Engine
Checks whether a local LLM fits on a GPU or Apple Silicon Mac with architecture-aware memory math.
Tested on Linux
Use it
Linux x86_64
Install fitllm-engine globally via npm
npm install -g fitllm-engine
Check that it works
fitllm --helpThe output should contain “exit codes: 0 fits”.
First thing to try
fitllm "Gemma 4 12b" --gpu "RTX 4090"Run fitllm "<model name>" --gpu "<GPU name>" to check whether a model fits your GPU.
What it does
- Check if a model fits your GPUTested
How to use it
- Run
fitllm "<model name>" --gpu "<GPU name>". - Read the verdict, which shows the GPU memory, quantization and context size.
- Run
- Rank models that fit your hardwareTested
How to use it
- Run
fitllm --top --gpu <GPU name>. - Read the ranked list, which shows memory use and maximum context for each model.
- Run
- Scan downloaded Ollama and LM Studio modelsNot shown working
- Detect your local GPU automaticallyNot tested
How far pickfit got
Linux x86_64
- Installs
- Starts
- 2 of 4 features work
- Works offline
macOS and Windows: the project says it runs there; pickfit hasn’t tried it.
Good for
- check one model against a GPU or Mac and return fit verdict
- rank supported local models that fit given hardware
- detect local GPUs and list what fits as JSON
How pickfit tested it
Each run starts from a clean machine and follows the steps above. How pickfit verifies
Linux x86_64
Its main function works
- Replayed from scratch
- Works offline
- Checked on 30 Oct 2026
- Version 2.22.0
Differences the test machine hit
- README: npm install fitllm-engine
In the test: npm install -g fitllm-engine puts fitllm on PATH via ~/.local prefix; README examples use npx fitllm - README: --version or --help check (only --help works)
In the test: use fitllm --help; fitllm --version is parsed as a model name and exits 2 with unknown model