llama.cpp Release b11372 Halves Indexer Score Memory

Core summary

One sentence to understand this update

The llama.cpp b11372 release incorporates an optimization, qwen4exp, that significantly reduces the memory consumption for indexer scores by half.

Impact & opportunity

What this could mean

This memory optimization can improve the efficiency and accessibility of llama.cpp for users with constrained hardware, enabling larger models or more concurrent operations.