b10267
ggml-org/llama.cppb10267Aug 4, 2026by github-actions[bot]same day as b10265
What changed since b10265
Internal refactor of speculative decoding init to cut duplication; no functional or API changes between b10265 and b10267.
Improvements
- Consolidated the enabled speculative decoding configuration handling in common_speculative_init, removing duplicated code paths.
- Code-only cleanup with existing speculative decoding server tests still passing, indicating behavior is unchanged.
Upgrade notes
This is a drop-in replacement: no CLI flags, config keys, or APIs were altered, and speculative decoding behaves as before. The older release was a ggml library sync, so the only delta here is internal refactoring. Standard rebuild and re-run of your speculative decoding tests is sufficient.
Analyzed Oct 31, 2026
Upgrading?
Release notes summary
Refactors the initialization code for speculative decoding to reduce code duplication.
Key Highlights
- Refactored common_speculative_init
- Reduced code duplication
- Existing tests still pass
Release Notes (excerpt)
<details open> speculative : refactor enabled configs common_speculative_init (#26510) This commit contains a suggestion to reduce some code duplication in common_speculative_init when adding the enabled speculative decoding configurations. No tests were added but the existing server tests still passes with this change: ```console $ ./tests.sh unit/test_speculative.py -v -x ``` </details> **Website:** - <https://llama.app> **macOS/iOS:** - macOS Apple Silicon (arm64)…
Published by ggml-org on GitHub.
Read the full release notesRelated Repositories
:electron: Build cross-platform desktop apps with JavaScript, HTML, and CSS
Godot Engine – Multi-platform 2D and 3D game engine
ClickHouse® is a real-time analytics database management system
Self-hosted game stream host for Moonlight.
DuckDB is a high-performance analytical database system