llama.cpp b11485 Fixes Jinja Chat Template Parsing for TranslateGemma
llama.cpp release b11485 introduces a fix for the Jinja template parser when processing prompt templates for TranslateGemma models. The update also adds warning logs when required parameters like source_lang_code or target_lang_code are missing. This bug fix ensures seamless local execution and chat template formatting for Google's new TranslateGemma translation models within llama.cpp. It prevents prompt rendering failures for users relying on llama.cpp for multilingual translation workflows. The change was merged via PR #30096 to address Jinja parsing issues specific to TranslateGemma chat format. Pre-built release binaries have been provided across macOS, Linux, Windows, Android, and Snapdragon platforms.
## BACKGROUND
llama.cpp is a popular open-source LLM inference framework written in C/C++ that enables high-performance local AI execution on consumer hardware. TranslateGemma is a specialized open translation model family built by Google on Gemma 3 to support translation across 55 languages. Jinja templates are used in llama.cpp to format structured conversational messages into the specific raw prompt strings expected by different model architectures.