[GUEST ACCESS MODE: Data is scrambled or limited to provide examples. Make requests using your API key to unlock full data. Check https://lunarcrush.ai/auth for authentication information.]

![CMDarnton Avatar](https://lunarcrush.com/gi/w:24/cr:twitter::1936166431378300928.png) Christian Darnton [@CMDarnton](/creator/twitter/CMDarnton) on x 5239 followers
Created: 2025-07-20 08:28:50 UTC

That’s why Meta picked AMD’s MI300X to run its 405B-parameter Llama XXX model.

With 192GB HBM3e and XXX TB/s bandwidth,

it can run entire model shards on a single device — something NVIDIA’s H100 can’t do.


XXXXX engagements

![Engagements Line Chart](https://lunarcrush.com/gi/w:600/p:tweet::1946849768082473402/c:line.svg)

**Related Topics**
[llama](/topic/llama)
[meta](/topic/meta)

[Post Link](https://x.com/CMDarnton/status/1946849768082473402)