Benchmark local LLMs for performance, quality, and RAM compatibility.
npx skills add https://github.com/MetriLLM/metrillm --skill metrillm
Please help me install this Agent Skill. Skill: metrillm Source: https://github.com/MetriLLM/metrillm/tree/main Command: npx skills add https://github.com/MetriLLM/metrillm --skill metrillm
This Skill enables users to evaluate the speed, quality, and RAM compatibility of local large language models (LLMs) on their hardware.
Run the metrillm bench --model your-model command to test your local LLM for speed and quality, then review the benchmarks directly.
High-intent search queries and answers about installing and using this skill.
Evaluate local LLM quality by testing reasoning, math, coding, instruction following, and multilingual capabilities using built-in benchmark prompts to assess deployment readiness and model accuracy.
Yes, you can test hardware compatibility by measuring RAM compatibility and speed metrics of specific local models to ensure they run efficiently on your machine before deployment.
Yes, you need ollama or lm-studio installed along with node to provide the local model environment required to execute performance and quality tests effectively.
The best way to compare local LLM speed and quality is to run comprehensive benchmark tests measuring both performance metrics like tokens per second and quality assessments like reasoning and coding accuracy.
lucasfthRun local LLM inference and quantize GGUF models with llama.cpp.
richardnguyen0715Deploy and serve LLMs with vLLM, OpenAI-compatible endpoints, and quantization.
ricableRun local GGUF model inference with Metal and CUDA acceleration.
MetriLLMBenchmarks local LLM models for speed and output quality.
