A 30B model that runs agents on a single consumer GPU is the open-weights stor… · @harden_369741 · TwKit

@harden_369741

A 30B model that runs agents on a single consumer GPU is the open-weights story I actually care about. I have watched open models win benchmarks and still lose the real world because nobody could run them where the work happens. If Glimmer really leads its size class on tool use while fitting on a Mac, the agent stack stops being a datacenter luxury. How long until the best agent on my laptop is not from a cloud API at all?

曝光 10 · 评论 0 · 点赞 0 · 书签 0