TGViewer
Aziz Mirzayev Aziz Mirzayev @pro_aim · 926 subscribers
Post #1705 259
DeepSeek KV cache'ni 437 barobar kichraytirdi. G'arb laboratoriyalari esa jimgina narxni tushirdi 🤔

insufferable.dev muallifining fikri: DeepSeek'ning MLA, CSA va FP4 caching kabi optimizatsiyalari ochiq e'lon qilingan. V4.1-Flash'da KV cache bir token uchun atigi 890 bayt.

Opus 5.5 va GPT-6.1 Sol ortiqcha shovqinsiz chiqdi, cache-read narxi esa 60% va 80% ga tushdi. Muallif buni xitoy g'oyalarini indamay o'zlashtirish deb hisoblaydi.
Coding agent'lar uchun bu juda muhim: uzun context ham arzonlashyapti =)

Nima deb o'ylaysizlar: GPU cheklovi xitoyliklarni kuchliroq qildimi?

Manba: https://insufferable.dev/posts/the-ai-race-just-got-awkward/
insufferable.dev The AI Race Just Got Awkward | insufferable.dev Funny how quiet everyone got.
More from @pro_aim
  1. Oct 3, 2026UzGeeks October2026 Design Day meetup! @vejon_uz
  2. Oct 3, 2026AI kod yozishni tezlashtirdi. Endi bottleneck biz, ya'ni PR review qiladigan odamlar 😄 Ma…
  3. Oct 3, 2026LLM'dan javobni matn qilib emas, video qilib so'rang! Karpathy'ning maslahati: oddiy matn…
  4. Oct 3, 202620 minuttan keyin boshlimiz.
  5. Oct 2, 2026Бекорга этмимандэ соат 3 дан кейин кофе ичиш керемас =) Боя Корасаройда аччик кофени 5 лар…
  6. Oct 2, 2026#839515. Оптимизировал наш sp-* флоу и замерил на этой же задаче, старые скилы против новы…
Threads Profile ViewerView any public Threads profile without an account.Open ThreadLook →Writing with AI? Make it sound human.Metric37 rewrites AI drafts so they read naturally. Free AI detector, 1,500 words free.Try Metric37 →