IBM Granite 4.2

Granite 4.2 30B is the flagship reasoning model in the Granite 4.2 family, designed for advanced coding, mathematics, tool calling, agentic workflows, and multilingual dialogue. It features 30B parameters with a 64-layer dense transformer using 4,096 hidden size, 32 attention heads, and 8 KV heads. The architecture employs Grouped Query Attention, RoPE with a 50M theta, and a 32,768-dimensional SwiGLU feed-forward network, with native reasoning and flexible thinking modes for balancing quality and latency. Supporting 128K native context with extension to 512K, it is optimized for complex long-context reasoning and enterprise applications.

Granite 4.2 8B is a small-size reasoning model designed for coding, mathematics, tool calling, agentic workflows, and multilingual dialogue. It features 8B parameters with a 40-layer dense transformer using 4,096 hidden size, 32 attention heads, and 8 KV heads. The architecture employs Grouped Query Attention, RoPE with a 10M theta, and a 12,800-dimensional SwiGLU feed-forward network, with native reasoning and flexible thinking modes for balancing quality and latency. Supporting 128K native context with extension to 512K, it is optimized for efficient long-context reasoning and enterprise applications.

Granite 4.2 3B is a compact reasoning model designed for coding, mathematics, tool calling, agentic workflows, and multilingual dialogue. It features 3B parameters with a 40-layer dense transformer using 2,560 hidden size, 40 attention heads, and 8 KV heads. The architecture employs Grouped Query Attention, RoPE with a 10M theta, and an 8,192-dimensional SwiGLU feed-forward network, with native reasoning and flexible thinking modes for balancing quality and latency. Supporting 128K native context with extension to 512K, it is optimized for efficient long-context reasoning and enterprise applications.