Qeravio
Canonical AI event

ggml updates IQ1_M to build prefix sums once per block.

IQ1_M now builds prefix sums once per block in ggml. The update supports multiple platforms including Ubuntu, Windows, and macOS.

22 Sept 20261 verified claims1 sources1 observations
What happened

The official source reports this update: b11108: ggml : IQ1_M build prefix sums once per block.

Why it matters

This official update documents a development concerning b11108: ggml : IQ1_M build prefix sums once per block. Its practical significance depends on the scope and evidence stated by the source.

What to watch next

Read the official source update and verify its stated scope, evidence and timing before acting on it.

Connected knowledge

Entities affected by this event

Llama Cppsubject
Continue this topic
infrastructureMetal gate mul_mm_id src1 precision change disables macOS Apple Silicon support.22 Sept 2026infrastructureJinja updates statement::execute and ::visit to use const22 Sept 2026infrastructureLlama-server now supports binding to multiple addresses.22 Sept 2026infrastructureHunyuanOCR now supports DFlash speculative decoding.22 Sept 2026infrastructureLlama.cpp adds MiMo-V2.6 support across multiple platforms.22 Sept 2026infrastructureLlama.app fixes Muse Glimmer tool-call parser error.22 Sept 2026
Evidence trail

Sources behind the event

llama.cpp official releasesb11108: ggml : IQ1_M build prefix sums once per block
Editorial presentation

Read the public story

Open story