Nicholas Dunzelman(@nicdunz):They shrunk a strong 27B model to 5.9 GB. 9× smaller keeping ~98% of original performance. “Ternary” compression stores model weights using only three values. That means powerful AI can run on much cheaper, smaller hardware.

Nicholas Dunzelman

@nicdunz

AI, models & creative tech · @krea_ai

加入 March 2023

601 正在关注 14.5K 粉丝

Nicholas Dunzelman@nicdunz

2026.09.18 03:01

They shrunk a strong 27B model to 5.9 GB. 9× smaller keeping ~98% of original performance. “Ternary” compression stores model weights using only three values. That means powerful AI can run on much cheaper, smaller hardware.

显示更多

转发到社区

热门用户

254.7K 粉丝

1.9M 粉丝

4.2M 粉丝

181.3K 粉丝

33.9K 粉丝

23.7K 粉丝

478.6K 粉丝

100.8K 粉丝

256.6K 粉丝

547.4K 粉丝

733.3K 粉丝

348.8K 粉丝

129.5K 粉丝

20.7K 粉丝

8.5K 粉丝