Get the latest tech news

Show HN: DeepSeek V3 in ~200 lines of pure Jax

DeepSeek V3 (MLA, DeepSeekMoE, aux-loss-free balancing, MTP) in 190 lines of pure JAX, trained on a MacBook Air. Write-up: vaibhaav.in/writing/deepseek-v3-jax - Vaibhaav-Tiwari/deepseek-v3-jax

None

Get the Android app

Or read this on Hacker News

Related news:

Developer cleverly shrank 2,600 lines of C into a six-line script but still ending up losing sleep

Linux 7.4 Merge Window Saw Nearly 20k Commits, Source Tree Approaching 42 Million Lines

With 86% of the words, you might understand only 38% of an anime's lines

« Thousands more flights canceled as Duffy warns the FAA's aging system could fail again

Uber drivers' information stolen in a third-party hack »