ggerganov/llama.cpp Released Version b12004
The llama.cpp project released version b12004, introducing a specific update to fix reasoning budget forcing on an end tag from the prompt. This maintenance update continues the frequent incremental improvements to the library. This release helps developers and AI practitioners maintain stable local LLM inference behavior, particularly for models that rely on structured reasoning budgets and prompt tags. Maintaining control over reasoning parameters ensures more predictable output and better integration with consumer hardware. The update specifically resolves issue #31374 regarding reasoning budget handling during prompt parsing. Precompiled binaries and source packages are provided across multiple platforms including macOS, Linux, Android, and Windows with various hardware backends.
## BACKGROUND
llama.cpp is a popular open-source software library created by Georgi Gerganov that enables efficient inference of large language models on consumer hardware. It serves as the core backend for numerous local AI tools and applications by running model weights directly via C/C++ implementations.