b9830

ggml-org/llama.cppb9830Jun 28, 2026by github-actions[bot]same day as b9829

What changed since b9829

low risk

b9830 exposes an --offline flag for `llama download` and fixes a use-after-free in the URL-task callback.

New features

  • The `llama download` command now accepts an `--offline` flag, letting scripts verify whether a model is already cached without making network requests.

Improvements

  • Reuses the existing offline flag plumbing in the common download path rather than introducing new behavior.

Bug fixes

  • Fixed a latent use-after-free where the block-scoped `first_path` was captured by reference in the URL-task `on_done` callback and invoked after its scope ended.

Upgrade notes

This is a small, drop-in update with no config or API changes required. Scripts that check model cache status can now pass `--offline` to `llama download` to avoid network access. The use-after-free fix is internal and needs no action beyond upgrading.

Analyzed Oct 31, 2026

Upgrading?

Release notes summary

Adds an offline mode to the `llama download` tool and fixes a critical use-after-free bug in the URL-task callback.

Key Highlights

  • Exposed the `--offline` flag for the `llama download` command
  • Fixed a use-after-free bug in the URL-task on_done callback
  • Allows scripts to check cache status without network requests

Release Notes (excerpt)

<details open> common : allow --offline in llama download (#25091) Expose the existing --offline flag to `llama download` so a script can run it to check whether a model is already cached and ready to be served without touching the network. Also fix a latent use-after-free in the URL-task on_done callback: first_path is block-scoped and was captured by reference, but invoked after the block ends. Signed-off-by: Adrien Gallouët <angt@huggingface.co> </details> **macOS/iOS:**…

Published by ggml-org on GitHub.

Read the full release notes

Related Repositories