SettledSafety
IBM releases LLMON preview models that mark prompt structure with special tokens to resist injection
IBM's Granite team released two experimental preview models, granite-4.1-3b-llmon-preview and granite-4.1-8b-llmon-preview, trained on LLMON, a markup language whose delimiters are reserved tokens. IBM says they avoid 47% more static prompt-injection attacks than the best frontier model it tested on the OPI benchmark. Adaptive multi-turn attacks are untested, and IBM says they are not for production.
- HEAT
- OUTLETS
- 0
- FIRST SEEN
- LAST SIGNAL
Primary source
Hugging Facehuggingface.co/blog/ibm-granite/llmon-an-llm-native-markup-language-now-in-graniteCoverage · 1 article, oldest first
Thursday, November 12, 2026
Hugging FacePrimary
LLMON: An LLM-Native Markup Language, now in Granite 4.1 3B and 8B