Granite 4.1 releases LLMON preview models 3B and 8B

Researchers have introduced LLMON, an LLM-native markup language designed to communicate structure and semantics explicitly to causal language models using reserved special-token delimiters. Alongside this new language, experimental preview models derived from Granite 4.1 base checkpoints—granite-4.1-3b-LLMON-preview and granite-4.1-8b-LLMON-preview—have been released, demonstrating near-zero attack success rates against static indirect prompt injections while maintaining general capability. Open article →