NHacker Next
  • new
  • past
  • show
  • ask
  • show
  • jobs
  • submit
▲From 300KB to 69KB per Token: How LLM Architectures Solve the KV Cache Problem (news.future-shock.ai)
24 points by future-shock-ai 3 days ago | 1 comment
az09mugen 49 minutes ago [-]
Unrelated, but 69KB is how much RAM Voyager 1 has.
Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact
Rendered at 18:58:13 GMT+0000 (Coordinated Universal Time) with Vercel.