signerless/llm-checker

LLM Checker is a CLI tool that analyzes your hardware and recommends which large language models you can run locally with Ollama integration.

The tool solves the problem of selecting appropriate LLM models for your specific hardware configuration. It scans your system to detect available compute resources—including Apple Silicon, NVIDIA CUDA, AMD ROCm, Intel Arc, and CPU options—then scores compatible models across four dimensions: Quality, Speed, Fit, and Context. The scoring engine weights these dimensions based on your use case and relies on a packaged multi-source registry containing exact installable artifacts from Hugging Face, Ollama, and GPT4All. Memory estimation uses a bytes-per-parameter formula validated against real Ollama sizes to ensure recommendations account for your actual VRAM constraints.

You should adopt this tool if you need to quickly identify which models your hardware can support without manual research across multiple sources. It suits developers building local AI applications who want deterministic, hardware-calibrated recommendations rather than guesswork. The tool ships as pure JavaScript with no native dependencies, running on any Node.js system. The README mentions comparison with llmfit but does not detail how the two differ, so no comparative assessment can be made here. The packaged catalog includes over two hundred models synced from Ollama and can refresh on demand, while the multi-source registry covers thousands of exact artifacts across three platforms.

The project maintains a synced SQLite catalog of models and provides live metrics showing token-per-second response speeds during model execution. It includes a Claude MCP integration for code assistance and ships with calibration fixtures and comprehensive documentation covering usage, advanced techniques, and technical reference material. The tool offers a verify command powered by ModelVet's structural verification, enabling validation of model integrity and policy compliance.