Even more important, subjectively, is that this model will run very well on Strix Halo (e.g. Framework Desktop), DGX Spark kinds of devices.
Looking forward to Unsloth dynamic mtp quants.
P.S. Looking at the HF release they already offer Q4_K_M and DFlash drafter for speculative decoding!
tosh 54 minutes ago [-]
open weights and
similar performance to deepseek v4, inkling at size of nemotron 3 super (!)
iraldir 49 minutes ago [-]
Amazing model at this size if true, that's quite crazy!
megavon 44 minutes ago [-]
This is INSANE. How did they do this?
eisokant 15 minutes ago [-]
"What we've done in this model is not necessarily add more intelligence, but improve the behaviors that lead to a more capable model: more verification, less taking things for granted, not declaring victory early, and being more persistent.”
Even more important, subjectively, is that this model will run very well on Strix Halo (e.g. Framework Desktop), DGX Spark kinds of devices. Looking forward to Unsloth dynamic mtp quants.
P.S. Looking at the HF release they already offer Q4_K_M and DFlash drafter for speculative decoding!
similar performance to deepseek v4, inkling at size of nemotron 3 super (!)
+
https://poolside.ai/assets/laguna/laguna-m1-xs2-technical-re...