b8162

The ggml-org team behind the massively popular llama.cpp project has released version b8162, addressing a significant server configuration bug that affected users across all supported platforms. The commit specifically fixes issue #19897 where the 'load-on-startup' setting in INI configuration files wasn't being properly respected by the server component, potentially causing inconsistent behavior when deploying llama.cpp in production environments. This release comes as llama.cpp continues to dominate the open-source LLM inference space with 96k GitHub stars and 15.1k forks, making even minor configuration issues impactful for thousands of developers and organizations.

The technical fix ensures that server instances now correctly honor the load-on-startup directive, which controls whether models are loaded into memory immediately when the server starts or deferred until first request. The release includes pre-compiled binaries for an extensive range of platforms including macOS (both Apple Silicon and Intel architectures), Linux (with CPU, Vulkan, and ROCm 7.2 backends), Windows (supporting CPU, CUDA 12/13, Vulkan, SYCL, and HIP), iOS via XCFramework, and multiple openEuler configurations. This cross-platform consistency is crucial for developers deploying llama.cpp across heterogeneous infrastructure, particularly as the project becomes increasingly central to enterprise AI deployments where configuration management and predictable startup behavior are essential requirements.