Post by sarah guo on X
sarah guo@saranormous
XI don’t think the US chip export controls are having their intended effect. Chinese model DeepSeek v3 very strong, and trained with OOM less money:
“DeepSeek-V3 requires only 2.788M H800 GPU hours for its full training”
(h800 is h100 with lower interchip bandwidth)

279 likes40 repliesPosted Dec 26, 2024