FitMyLLM

Running LLMs on NVIDIA CMP 40HX: What You Can Run

The NVIDIA CMP 40HX packs 8GB of GDDR6 with 448 GB/s of memory bandwidth, making it an entry-level option for running local AI models. Here is which models fit, how fast they run, and what quantisation gives the best balance of quality and speed.

Already have it? Find the best model for it. Still choosing? See what a budget buys. Or the full specifications.

Loading the model tables…

Running LLMs on NVIDIA CMP 40HX: Complete Guide

The NVIDIA CMP 40HX has 8GB VRAM and 448 GB/s memory bandwidth, making it an entry-level option for running local AI models. This guide covers which LLMs fit, expected tok/s performance, and recommended settings.

See the full NVIDIA CMP 40HX specs page for detailed specifications, all compatible models, and speed estimates.