llama.cpp b9632 adds Jinja filter aliases for better LLM templating

The open-source llama.cpp project, known for enabling local inference of large language models, has released version b9632. This update introduces Jinja filter aliases for the count/d/e filters, enhancing the flexibility of prompt templating. Developers can now use these aliases to manipulate variable counts and formatting more intuitively when constructing prompts for models like Llama and Mistral.

The release includes pre-built binaries for multiple platforms: macOS (Apple Silicon arm64, Intel x64), Linux (x64/arm64/s390x with CPU, Vulkan, ROCm 7.2, OpenVINO, SYCL), Windows (x64/arm64 with CPU, CUDA 12/13, Vulkan, SYCL, HIP), and Android arm64. All builds are signed with GitHub's verified signature. This new feature builds on llama.cpp's commitment to making LLM deployment accessible on consumer hardware, from laptops to gaming GPUs.