IBM Rolls Out Granite 4.2 Open-Weight Models in 3B, 8B and 30B Sizes for Local Deployment

IBM has released the newest models in its family of open-weight large language models, Granite 4.2, available in 3B, 8B and 30B parameter sizes. All can be downloaded and self-hosted. The release focuses on agentic capability and predictable enterprise deployment, riding a wave of interest in local LLMs.

Three sizes and agentic capability

Like previous versions, Granite 4.2 takes a decoder-only approach, and every variant offers a 128,000-token context window natively. The 8B and 30B variants (not the 3B one) go through an agentic reinforcement learning block, trained for capabilities such as using the terminal, searching the web or using external tools. The 3B model supports tools too, but without the same level of specialized training.

IBM described Granite 4.2 as the reasoning-focused release of the Granite language-model family. Reasoning here does not mean conscious understanding; it refers to functional reasoning, particularly via chain-of-thought, carrying intermediate results forward through multiple steps. For the user, this can mean more rigorous and accurate responses in some cases, but often slower response times and higher compute demands.

The local deployment trend

The Granite family rarely grabs headlines for speed or aggressive innovation. Its pitch is predictable deployment, which is the priority many enterprises look for. With ongoing discussion about the cost and compute crunch around frontier cloud models from the likes of Anthropic and OpenAI, individual developers and enterprises have been exploring local models as cheaper alternatives. Interest in model routers, tools that interpret user prompts and route them to appropriately scoped models, has also grown.

Models like this are popular with hobbyists, AI researchers and individual developers because they can be tinkered with on local hardware without per-token API fees.

Source: Ars Technica