Skip to content

← Back to Freedom.Tech Back

All llama.cpp releasesAll versions →

Post to Nostr Post to X

Release Thu, Oct 15, 2026 1 min read

llama.cpp b11640

At a glance

Vulkan backend now uses L2-friendly workgroup scheduling for matrix multiplication, improving GPU cache efficiency and throughput on compatible hardware

Original release notes

vulkan: Use L2-friendly workgroup scheduling for cm2 mulmat (#30215)

Website: -

Attestations: -

UI:

UI

Release context

Release on GitHub · Diff vs previous · All llama.cpp releases · Project homepage

Previous release: Previous on Wed, Oct 14, 2026 · 1 day apart

Filed under

AI All llama.cpp releases →#ai #release

More in AI

AI release Oct 22, 2026

llama.cpp b11808

AI release Oct 21, 2026

llama.cpp b11803

AI release Oct 22, 2026

Claude Agent Sdk Python 0.2.172

Get this in your inbox

The week in freedom tech, synthesized into one Sunday email.

Powered by Foundation

About Sources All briefs RSS Nostr X