llama.cpp Release b9659 Fixes Miscounting of n_tokens in mtmd (#24656)
English summary
The llama.cpp project released tag b9659, which includes a bug fix for the mtmd component that was miscounting n_tokens (PR #24656). This release also provides pre-built binaries for a wide range of platforms, including macOS (ARM64, Intel), Linux (x64, ARM64, s390x with Vulkan, ROCm, OpenVINO, SYCL), Android (ARM64), and Windows (x64, ARM64 with CUDA 12/13, Vulkan, SYCL, HIP). Notably, the macOS Apple Silicon build with KleidiAI enabled is marked as disabled, while the iOS XCFramework artifact is available.
Chinese summary
llama.cpp 发布标签 b9659，包含对 mtmd 组件中 n_tokens 计数错误的修复（PR #24656）。该版本同时提供了面向多平台的预编译二进制文件，涵盖 macOS（ARM64、Intel）、Linux（x64、ARM64、s390x，支持 Vulkan、ROCm、OpenVINO、SYCL）、Android（ARM64）和 Windows（x64、ARM64，支持 CUDA 12/13、Vulkan、SYCL、HIP）。其中，启用 KleidiAI 的 macOS Apple Silicon 构建被标记为禁用，而 iOS XCFramework 产物可用。
Key points
Fixes a miscounting of n_tokens in the mtmd component of llama.cpp (PR #24656).
修复 llama.cpp 中 mtmd 组件对 n_tokens 的计数错误（PR #24656）。
Release b9659 includes the fix and provides pre-built binaries for macOS, Linux, Android, and Windows, covering CPU, GPU, and various accelerator backends.
b9659 版本包含此修复，并提供 macOS、Linux、Android 和 Windows 的预构建二进制文件，涵盖 CPU、GPU 及多种加速器后端。
The macOS Apple Silicon build with KleidiAI support is disabled in this release.
此次发布中，启用 KleidiAI 的 macOS Apple Silicon 构建处于禁用状态。