b10435

ggml-org/llama.cppb10435Aug 14, 2026by github-actions[bot]same day as b10434

What changed since b10434

low risk

Single jinja fix removes quadratic cost in gather_string_parts, speeding up template string processing.

Improvements

  • Jinja template string gathering now runs in linear rather than quadratic time, reducing CPU cost for large or complex templates.
  • Memory efficiency improves for template rendering since intermediate string parts are handled more cheaply.

Bug fixes

  • Fixed a quadratic-time bottleneck in gather_string_parts that could make template evaluation scale poorly with input size.

Upgrade notes

This is a drop-in patch release with no config, API, or template syntax changes; existing chat templates and server flags continue to work as-is. Upgrading is worthwhile if you use heavy jinja templates or long prompts, where the performance gain is most visible.

Analyzed Oct 24, 2026

Upgrading?

Release notes summary

Performance optimization release that fixes quadratic cost issues in jinja template string processing, improving efficiency for template operations.

Key Highlights

  • Fixed quadratic cost in gather_string_parts
  • Performance optimization for jinja templates
  • Memory efficiency improvements

New Features

  • Optimized string gathering in jinja templates

Release Notes (excerpt)

<details open> jinja : fix quadratic cost in gather_string_parts (#27034) * jinja : fix quadratic cost in gather_string_parts * fix some comments * remove test </details> **Website:** - <https://llama.app> **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)…

Published by ggml-org on GitHub.

Read the full release notes

Related Repositories