b10248

ggml-org/llama.cppb10248Aug 4, 2026by github-actions[bot]same day as b10247

What changed since b10247

low risk

b10248 adds validation of default special token IDs during vocabulary loading, with no other changes noted.

New features

  • Vocabulary loading now validates default special token IDs for consistency.

Improvements

  • Added integrity checks on default special token IDs to catch malformed or inconsistent vocab definitions early.

Upgrade notes

This is a single-commit build bump with no config or API changes; binaries can be swapped in place. Models with unusual or hand-edited tokenizer metadata may now surface validation warnings or errors at load time that were previously silent. The MoE split-graph allocation fix from b10247 remains included.

Analyzed Nov 1, 2026

Upgrading?

Release notes summary

Validates default special token IDs within the vocabulary system to ensure consistency.

Key Highlights

  • Validation of default special token IDs
  • Enhanced vocabulary integrity checks

Release Notes (excerpt)

<details open> vocab : validate default special token ids (#26506) </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) - Ubuntu x64 (ROCm 7.2) - Ubuntu x64 (OpenVINO) - Ubuntu x64 (SYCL FP32) - Ubuntu x64 (SYCL FP16) **Android:**…

Published by ggml-org on GitHub.

Read the full release notes

Related Repositories