---
title: "Using the formulation, I calculated theoretical upper bounds for all the models on Hugging Face..."
date: 2026-07-17
canonical: https://solmaz.io/x/2078193848896065558/
x_url: https://x.com/onusoz/status/2078193848896065558
license: CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/)
---

Using the formulation, I calculated theoretical upper bounds for all the models on Hugging Face that have downloads over 100k, matched against a database of consumer GPUs.

Here is an example of Qwen/Gemma NVFP4 quantizations and antirez/ds4 on the DGX Spark and some other 128 GB Mac configurations

https://osolmaz-local-frontier.hf.space/compare?models=nvidia%2FQwen3.6-35B-A3B-NVFP4,nvidia%2FGemma-4-26B-A4B-NVFP4,antirez%2Fdeepseek-v4-gguf&hardware=nvidia-dgx-spark,apple-macbook-pro-16-m5-max-128gb,apple-mac-studio-m1-ultra-128gb

*Part 2/2 of a thread; root: https://solmaz.io/x/2078193844966080981/*