# Why does Minimax H3 Image even run (and fast!) on my 8GB VRAM/32GB RAM system despite the entire stack being about ~50GB (int8 pruned)? — retained record

> Retained reddit record from reddit:StableDiffusion.

Canonical: https://highsignal.app/data/records/04d2a2917271782e

[← sources](https://highsignal.app/data)


source: [reddit](https://highsignal.app/data/reddit) (reddit:StableDiffusion)  published 2026-11-13 20:08 ingested 2026-11-14 02:38

[original source ↗](https://www.reddit.com/r/StableDiffusion/comments/1xvi9ko/why_does_minimax_h3_image_even_run_and_fast_on_my/)

## retained text



```
i2i, 2 1024x1024 reference photos. 

My system is RTX 2070 8GB (Turing), 32GB DDR4 (2666Mhz), Ryzen 1700 (1st gen). Ubuntu 24.02 server.

\~ 57 seconds per image! (ComfyUI)

This is MUCH faster than my Krea2, Flux2 Klein, or Qwen 2.1 edit workflows! It's about the same as m Fibo2 edit workflow, but Fibo2 is **smol**!

My question is, how does this even not OOM? Correct me if I'm wrong, but don't ALL the models need to load entirely (including that stupidly massive 32b text encoder) into memory? 
```



document fetched 2026-11-14 00:31 · canonical https://reddit.com/r/StableDiffusion/comments/1xvi9ko/why_does_minimax_h3_image_even_run_and_fast_on_my
