llm-fit

Can my laptop run this model?

llm-fit gives instant local LLM fit + speed estimates from your hardware profile, then suggests exact run commands.

Why

People keep asking:

• Can I run 7B / 14B locally?
• How fast will it be on my machine?
• Which command should I use right now?

This tool answers in seconds.

Quick start

cd llm-fit
npm run cli -- --ram=16 --vram=8 --cpu=mid --platform=apple-silicon

Web demo (local)

npm start
# open http://localhost:4173

CLI output example

✅ Good fit  Qwen2.5 7B Q4
   ~33.6 tok/s | min VRAM 6GB | cmd: ollama run qwen2.5:7b

Flags

• --ram=16
• --vram=8
• --cpu=low|mid|high
• --platform=apple-silicon|nvidia|amd|cpu

How we differ from ollama-benchmark

• Built for real laptops, not lab servers – llm-fit focuses on what actually runs well on common Mac/PC hardware (RAM/VRAM tiers, CPU classes), instead of abstract throughput numbers from big server GPUs.
• Opinionated presets over config soup – instead of making you tune dozens of knobs, llm-fit gives curated presets for typical machines (e.g., "8GB MacBook Air", "32GB workstation") and recommends models that fit.
• End-user workflow first – the CLI and docs are written for people trying to get work done (chat, coding, local agents), not just benchmark graphs; every command is meant to be copy–paste runnable.

Next milestones

• Auto-detect machine hardware
• Pull benchmark data from community submissions
• Add GGUF quant-specific estimates (Q2_K/Q4_K_M/Q8_0)
• One-click export for Ollama / vLLM / llama.cpp

License

MIT

Architecture Overview

This project follows a modular structure with clear separation between interface, execution logic, and outputs/artifacts. The exact implementation details vary by repository, but the intent is to keep core logic testable and easy to extend.

Project Structure

.
├─ src/            # Core source code (if present)
├─ public/         # Static assets / UI resources (if present)
├─ docs/           # Documentation and notes (if present)
├─ scripts/        # Utility scripts (if present)
├─ test/           # Tests (if present)
└─ README.md       # Project overview

Folder names vary by project; this section describes the intended organization pattern.

Quick Start

1. Clone the repository
2. Install dependencies (if any)
3. Run the project using its local start/build instructions

If this repo is a library or static project, refer to scripts/config files for exact commands.

Current Scope

This repository reflects the project’s current implementation and active direction. Planned improvements are tracked through issues/commits and may evolve over time.