IBM's new Granite 4.2 models ride the wave of interest in local LLMs

AI models flub these intelligence tests. Can you fare any better?

Puzzles and games have been central to AI development since the very beginning. Just as we humans like to test our smarts with crosswords or logic puzzles, developers can test how far models have advanced with a gaming gauntlet.

Brief by WorldPing · Original reporting by MIT Tech Review