# Is it silly to get a 64GB Strix Halo (Framework Desktop) ~$2000? — retained record

> Retained reddit record from reddit:LocalLLaMA.

Canonical: https://highsignal.app/data/records/c4a1f0ac8b7f1007

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


source: [reddit](https://highsignal.app/data/reddit) (reddit:LocalLLaMA)  published 2026-09-01 17:12 ingested 2026-09-02 02:41

[original source ↗](https://www.reddit.com/r/LocalLLaMA/comments/1w4iaf9/is_it_silly_to_get_a_64gb_strix_halo_framework/)

## retained text



```
Hi! I've been considering getting a local AI station for video generation and light coding (I have coding AI subscription from work for heavyweight). My intended models are probably Minimax-H3 and Qwen 3.8 27b. 

I see many people recommending as much RAM as possible when you buy, but I feel like 64GB of unified memory fits my needs well - runs H3 and Qwen 3.8 27B with a lot of headroom for context. Is there a reason I should spend $1500 more for 128GB? Do you foresee video/small coding models getting inflated in size in the future? Also open to good alternatives to the Strix/Framework Desktop. Thanks a lot!!
```



document fetched 2026-09-02 00:43 · canonical https://reddit.com/r/LocalLLaMA/comments/1w4iaf9/is_it_silly_to_get_a_64gb_strix_halo_framework
