IBM's Granite LLMON Models Block 47% More Prompt Injections
IBM's Granite-4.1 LLMON previews avoided 47% more prompt-injection attacks than the best frontier model tested, cutting attack success rates to 0.02%–5.6%.
Topic
Llmon
Topic
Topic
Topic
IBM's Granite-4.1 LLMON previews avoided 47% more prompt-injection attacks than the best frontier model tested, cutting attack success rates to 0.02%–5.6%.