IBM has released the Granite 4.2 family of open-source reasoning models, featuring flagship 30B, mid-size 8B, and compact 3B variants. These models leverage native chain-of-thought capabilities to improve performance on math, coding, and complex multi-step problems.

  • All three models support flexible thinking modes (full, non-thinking, low-effort) to balance depth versus latency.
  • The architecture includes Grouped Query Attention, Rotary Position Embedding, and a 512K context window.
  • Models utilize reasoning-augmented tool calling for more accurate function invocation.
  • The family is Apache 2.0 licensed for commercial and research use.

The release provides users with open-weight options that allow switching between thinking modes within a single model to optimize for specific workflow requirements.