AINewsnow

128k context trên desktop là ảo tưởng: KV cache ăn hết RAM

This story is from 2026-09-22. It is preserved in the archive; the latest stories are on the live feed.

Originally published on NextFuture Bạn set num_ctx lên 128k cho con model 8B đang chạy ngon lành, rồi máy bắt đầu swap. Weights vẫn 16 GiB như cũ — thứ vừa ăn hết RAM là KV cache. Phép tính công khai và rất ngắn KV cache giữ hai tensor, keys và values, cho mỗi layer, mỗi KV head, mỗi token. Theo bà…

Read the full story at DEV Community — AI ↗

Timeline · 1 report

  1. 2026-09-22 23:00 · DEV Community — AI
    128k context trên desktop là ảo tưởng: KV cache ăn hết RAM

More stories

  1. GPT-6 Sol and Luna now available on AI Gateway — Vercel Blog
  2. Alibaba Unveils New AI Chip, Outlines Plan for Larger Model — Wall Street Journal Technology
  3. Anthropic launches Claude Opus 5.5 with stricter safeguards for cybersecurity — The Verge AI
  4. Amazon blocks Meta’s Muse AI agent — The Verge AI
  5. Moonshot’s Kimi K3 lands on Amazon in key test for Chinese open-source AI revenue — South China Morning Post Tech
  6. Trump says AI will be renamed 'super intelligence' in all US documents — The Hill Technology
  7. How Benchling secured multi-tenant AI agents with Amazon Bedrock AgentCore — AWS Machine Learning Blog
  8. Meet the Data Agent in ChatGPT Work — OpenAI YouTube

Get the daily brief of stories like this at 6:30 every morning →