ggerganov/llama.cpp Released b11798 with Jinja Hex Escapes
Llama.cpp release b11798 introduces support for hexadecimal string escapes within Jinja templates. It also provides standard pre-built multi-platform binary assets across macOS, Linux, Windows, and mobile environments. This update improves prompt template rendering compliance and flexibility when interacting with large language models that rely on Jinja formatting. It ensures that developers using local inference pipelines have up-to-date and reliable cross-platform builds. The release addresses pull request #30929, enabling hexadecimal character escapes in template strings. Additionally, other experimental features like KleidiAI support on Apple Silicon remain temporarily disabled pending further integration work.
## BACKGROUND
llama.cpp is a widely used open-source software library designed to perform efficient inference on large language models like Meta's Llama on consumer hardware. Jinja templates are frequently utilized by modern language models to format structured conversations, system prompts, and tool calls into a unified token sequence.