b9761

ggml-org/llama.cppb9761Jun 22, 2026by github-actions[bot]same day as b9760

What changed since b9760

medium risk

Server moves model downloads to a child process and adds real-time progress via /models/sse.

New features

  • Added a /models/sse endpoint for real-time model load and download progress tracking.
  • Model downloading in the router server now runs in a dedicated child process.

Improvements

  • Improved server responsiveness by isolating model downloads from the main server process.
  • Updated server documentation for the new download and progress-tracking behavior.
  • Added throttling around progress reporting to avoid excessive updates.

Bug fixes

  • Fixed test cases affected by the new download process and shorter test timeout.
  • Addressed cleanup and stability issues found during the download-process refactor.

Upgrade notes

No configuration migration is indicated, but clients that monitor model downloads should switch to or validate the new /models/sse progress endpoint. If the router server sits behind a proxy, ensure SSE connections and download timeouts are configured appropriately.

Analyzed Nov 1, 2026

Upgrading?

Release notes summary

Moved model downloading to a dedicated child process to allow for real-time progress tracking via SSE endpoint.

Key Highlights

  • Move model download to dedicated process
  • Real-time progress tracking via /models/sse
  • Improve server responsiveness

Release Notes (excerpt)

<details open> server: (router) move model downloading to dedicated process (#24834) * server: real-time model load progress tracking via /models/sse * update docs * server: move model download to child process * rm unused * fix most problems * clean up * nit fixes * fix test case * do not detact() thread * shorter MODEL_DOWNLOAD_TIMEOUT in test * throttle </details> **macOS/iOS:** - macOS Apple Silicon (arm64) - macOS Apple Silicon (arm64, KleidiAI enabled) DISABLED…

Published by ggml-org on GitHub.

Read the full release notes

Related Repositories