AnalysisAI ModelsNovember 18, 2026

Tencent's 220B Hy4-flash squeezed into 49 GB GGUF quants

Read original source →huggingface.co

Community quants of Tencent's 220B-parameter Mixture-of-Experts Hy4-flash now fit in 49 GB of weights, with experts at ~2.3 bits and everything else at Q8_0. The Engram n-gram table sits in a separate file kept on SSD, and one user ran an 85 GB IQ2 build on an RTX 3060 12 GB plus 32 GB RAM.

How this story unfolded

1 day · 0 reports · 3 community posts · from Nov 18

  1. Nov 18
  2. Nov 19

More stories today

Open the live feed