ggerganov/llama.cpp Releases b12019 with Qwen3-Coder Think Block Support
llama.cpp release b12019 introduces updates for handling think blocks in qwen3-coder models alongside standard binary builds for multiple operating systems. Specifically, it updates chat formatting to strip or accept think blocks when reasoning_format is set to none. This update ensures better compatibility with reasoning-capable coding models running locally. Users can control output formatting more precisely when interacting with models that generate internal reasoning steps. The release includes precompiled binaries and libraries for a wide range of platforms, including macOS Apple Silicon, Linux with various acceleration backends like CUDA and Vulkan, Android, and Windows. It also defines both qwen3-coder think block variants up front to improve chat handling.
## BACKGROUND
llama.cpp is an open-source software library that performs inference on various large language models and serves as a core component for many local AI tools. It is co-developed alongside the GGML project, which provides a general-purpose tensor library designed for efficient machine learning workloads.