llama.cpp b10088 fixes DeepSeek4 APE tensor operation bug

llama.cpp, the widely-used open-source project (121k GitHub stars) for running LLMs locally on consumer hardware, has released version b10088. The update is primarily a bugfix targeting the DeepSeek4 APE tensor operation, referenced in pull request #25945. This fix addresses a specific issue in the 'llama-arch' component, which handles model architecture support. For users running DeepSeek4 models—a popular architecture for reasoning and coding tasks—this patch ensures correct tensor computations, preventing potential crashes or incorrect outputs.

The release bundles pre-compiled binaries across an extensive range of platforms: macOS (Apple Silicon ARM64 with optional KleidiAI, Intel x64, iOS XCFramework), Linux (x64/ARM64/s390x with CPU, Vulkan, ROCm 7.2, OpenVINO, SYCL FP16/32), Windows (x64/ARM64 with CPU, OpenCL Adreno, CUDA 12.4/13.3, Vulkan, HIP), and Android (ARM64). The wide support underscores llama.cpp's role as the go-to framework for local inference, and this fix specifically improves reliability for DeepSeek model users.