b9722

ggml-org/llama.cppb9722Jun 19, 2026by github-actions[bot]same day as b9721

What changed since b9721

low risk

b9722 is a small follow-up to the ggml sync in b9721, fixing an unbound n_discard value in server context shifting.

Bug fixes

  • Server no longer uses an unbound n_discard value when shifting context, preventing incorrect context handling.

Upgrade notes

This is a patch-level bump with no config or API changes; the only code change is in the server's context-shift logic. Rebuild or pull the new binaries if you run llama-server and rely on context shifting. The ggml core sync from b9721 carries over unchanged.

Analyzed Oct 24, 2026

Upgrading?

Release notes summary

This release addresses a bug in the server component where non-bound `n_discard` values caused issues during context shifting. It also provides standard binary distributions for macOS, Linux, Android, Windows, and openEuler across various hardware architectures and backend support.

Key Highlights

  • Server fix for unbound n_discard value during context shifting
  • macOS/iOS binaries available for Apple Silicon, Intel, and XCFramework
  • Linux binaries supporting Vulkan, ROCm 7.2, OpenVINO, and SYCL
  • Windows binaries supporting CUDA 12.4 and 13.3

New Features

  • Server: fix non-bound n_discard value (ctx shifting)

Release Notes (excerpt)

<details open> server: fix non-bound n_discard value (ctx shifting) (#24786) * server: fix non-bound n_discard value * Update tools/server/server-context.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> </details> **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)…

Published by ggml-org on GitHub.

Read the full release notes

Related Repositories