llama.cpp b9273 ships critical server fix for multi-process routing

The llama.cpp project, a cornerstone of local AI inference with over 112,000 stars on GitHub, has released version b9273. This patch addresses a specific issue in the server component: when running under a unified binary configuration, the router was failing to re-inject subcommands when it spawned child processes. The fix, referenced in pull request #23442, ensures that the server correctly propagates subcommands across process boundaries, improving reliability for multi-process deployments.

This release continues llama.cpp's tradition of broad platform support. Binaries are now available for macOS (Apple Silicon with and without KleidiAI, Intel x64, iOS XCFramework), Linux (x64, arm64, s390x, plus GPU variants for Vulkan, ROCm 7.2, OpenVINO, SYCL FP32/FP16), Android (arm64), Windows (x64 and arm64 CPU, CUDA 12/13, Vulkan, SYCL, HIP), and openEuler (x86 and aarch64 with Ascend hardware support). Users can download the appropriate asset directly from the release page.