b10243

ggml-org/llama.cppb10243Aug 3, 2026by github-actions[bot]same day as b10242

What changed since b10242

low risk

b10243 makes a single change: indexer cache is now allocated only in 'full' indexer layers, cutting memory use.

Improvements

  • Indexer cache is now allocated only for 'full' indexer layers instead of all layers, lowering memory consumption.
  • Reduced per-model memory footprint for architectures that use indexer layers, leaving more headroom for context and batch size.

Upgrade notes

This is a narrow follow-up to b10242's backend penalties sampler work and requires no config or API changes. Users running indexer-based models should see lower memory usage after upgrading; no migration steps are needed.

Analyzed Nov 1, 2026

Upgrading?

Release notes summary

Optimized memory allocation by restricting indexer cache creation to specific 'full' indexer layers to reduce memory usage.

Key Highlights

  • Allocate indexer cache only in 'full' indexer layers
  • Memory optimization for indexer cache allocation

New Features

  • Memory optimization for indexer cache

Release Notes (excerpt)

<details open> llama : allocate indexer cache only in "full" indexer layers (#26474) Co-authored-by: Stanisław Szymczyk <sszymczy@gmail.com> </details> **Website:** - <https://llama.app> **macOS/iOS:** - macOS Apple Silicon (arm64) - macOS Apple Silicon (arm64, KleidiAI enabled) DISABLED - macOS Intel (x64) - iOS XCFramework **Linux:** - Ubuntu x64 (CPU) - Ubuntu arm64 (CPU) - Ubuntu s390x (CPU) - Ubuntu x64 (Vulkan) - Ubuntu arm64 (Vulkan) - Ubuntu x64 (ROCm 7.2) - Ubuntu x64 (OpenVINO)…

Published by ggml-org on GitHub.

Read the full release notes

Related Repositories