Re:GLM-4.7-Flash
by omgwin on Tuesday January 20, 2026 @12:11AM
(#46688144)
It's been mentioned that this model is MLA capable, but it seems like the default vLLM params don't use MLA. Seeing ~0.91MB KV Footprint per token right now. Are you getting MLA to work?
[ Parent
| on HN ]
There are no comments on this comment yet.