b9753

ggml-org/llama.cppb9753Jun 21, 2026by github-actions[bot]same day as b9752

What changed since b9752

low risk

Server progress reporting for loading spec models is fixed and now exposes a staged loading list.

New features

  • Server now exposes a 'stages' list that tracks model loading progress in discrete steps.

Improvements

  • Progress reporting during speculative model loading is more accurate and no longer misreports state.
  • Loading progress is now structured around named stages, making it easier to surface in server responses.

Bug fixes

  • Fixed incorrect progress reporting when loading spec models in the server.

Upgrade notes

This is a small server-side patch on top of the batch-construction refactor in b9752, so no config or API migration is required. Clients that parse progress output may see additional stage information, but existing fields remain usable. If you skipped b9752, note that it also reworked batch handling and added abort_all_slots.

Analyzed Nov 1, 2026

Upgrading?

Release notes summary

Fixed progress reporting for loading specification models and introduced a list of stages to track the loading process more accurately.

Key Highlights

  • Fix report progress for loading spec models
  • Add 'stages' list for tracking
  • Improve server progress reporting

Release Notes (excerpt)

<details open> server: fix report progress for loading spec models, add "stages" list (#24870) * server: fix report progress for loading spec models, add "stages" list * improve * nits * nits 2 </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) - Ubuntu s390x (CPU) - Ubuntu x64 (Vulkan) - Ubuntu arm64 (Vulkan) - Ubuntu x64 (ROCm 7.2)…

Published by ggml-org on GitHub.

Read the full release notes

Related Repositories