AINewsnow

How to Run Qwen3.8-Flash-Next Locally: 39 tok/s on 24GB of VRAM (Plus 96GB RAM)

Qwen3.8-Flash-Next only activates 6B parameters per token. Then you open Unsloth's GGUF repo and see that the smallest file, a 1-bit quant, is 66.4 GB, and the UD-Q4_K_XL that everyone defaults to is 111 GB. "Flash" describes how fast it decodes. It says nothing about how small it is. If you want t…

Read the full story at DEV Community — Machine Learning ↗

Timeline · 1 report

  1. 2026-11-08 21:00 · DEV Community — Machine Learning
    How to Run Qwen3.8-Flash-Next Locally: 39 tok/s on 24GB of VRAM (Plus 96GB RAM)

More stories

  1. Anthropic launches Claude Fable 5.2 and Mythos 5.2 models — Anthropic (X)
  2. Google adds TPU 8tx and 8i Max to eighth-generation AI chip lineup — Google: The Keyword
  3. Meta launches Muse Ember, a low-cost AI model for repetitive agent tasks — TechCrunch AI
  4. Australia to Adopt Mandatory AI Safeguards and Create AI Safety Regulator — Barron's
  5. Nvidia board probes CEO Jensen Huang over affair with alleged Chinese spy — Axios AI+
  6. Did Claude Code’s ‘don’t ask again’ option accidentally expose your enterprise's passwords and keys? — VentureBeat AI
  7. Kuken Industries uses OpenAI API to search 70 years of cooling tower drawings — OpenAI News
  8. Anthropic agrees $3.4bn deal for 1.7GW of gas generators to power data centres — Bloomberg AI

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