Original post unavailable.
Sentiment
Many users expressed excitement about Tobi Lutke hitting 40 TPS on GLM-52 with NVIDIA DGX because it offers strong localized inference density for medical AI workloads and reflects an impressive development pace.
Pos
100.0%
Neg
0.0%
3 comments with sentiment.
Cluster Engagement
Digg Deeper
No Digg Deeper questions have been answered for this story yet.
Related links
Posts from X
Most Activity
Most Activity
VIEWS4.5KREPLIES4
40 TPS on glm-52 (nvfp4+mtp)
Thank you @nvidia for the hookup on this DGX Workstation. This will crunch a *lot* of high quality tokens here! This thing is a total beast.
BOOKMARKS2LIKES34RETWEETS1
It’s building a liquid implementation from scratch overnight against http://gitHub.com/Shopity/liquid-spec
40 TPS on glm-52 (nvfp4+mtp)
@toddybayern @tobi I request that @tobi rebrands Shopify to Shopity immediately
@tobi how much is the budget?
@tobi how much wat it consumes? and what about the noise?
@tobi shopity would actually be a cool company name hahaha
@tobi GLM-52 at 40 TPS via NVFP4 + MTP? I would love to test that inference density on my medical AI workloads. Excellent localized compute.
@tobi "Overnight from scratch" means someone is very proud, very tired, and already thinking about the next sprint. That's a ridiculous pace to hit. Huge congrats.
@tobi Without the typo: https://github.com/Shopify/liquid-spec