IBM has launched Granite 4.2, the latest iteration of its open-weight large language models designed for local deployment. The release includes 3B, 8B, and 30B parameter variants, all featuring a 128,000-token context window and a decoder-only architecture.

The 8B and 30B models underwent agentic reinforcement learning, enabling advanced capabilities such as terminal usage, web searching, and external tool integration. The 3B model supports tools but lacks the same specialized training.

IBM describes this as the "reasoning-focused release" of the family. In AI terminology, reasoning refers to functional chain-of-thought processing — carrying intermediate results through multiple steps — rather than human-like conscious understanding.