llama.cpp PR drops kq_mask for QSA selection, saving gigabytes of VRAM on Qwen Flash Next
jacek2023 · reddit · 2026-10-24
A Reddit post highlights llama.cpp PR #30965 by sanmai: the qwen4exp change builds the QSA selection without the kqmask, reportedly saving gigabytes of VRAM for Qwen Flash Next. The PR has 9 comments, 8 reviews, +28/-14 lines across 9 commits, and is still under review.
More from Infra
- Baseten unveils Project Beacon to build safety tooling into the inference layer for open models — baseten · 2026-10-24
- Miles v0.1.3 cuts first RL weight update from 99.5s to 3.9s, shares GPUs across train/rollout — hsu_byron · 2026-10-24
- 98%+ Tensor Core utilization shown in LLM training run, drawing admiration — gharik · 2026-10-24
- We spent millions on AI apps with zero revenue — but our old GPUs are now worth more than we paid — stephen280up · 2026-10-24
- Polymarket puts 24% odds on AI data centers launching to space by end of next year — Polymarket · 2026-10-24
- Polymarket jokes wolves are entering Finnish villages after Google cleared forest for AI data centers — Polymarket · 2026-10-24