Skip to content
AI Radar by Wagner Solution
CS
Menu
worth noting New models

IBM has released Granite 4.2 open-weight models focused on reasoning and agentic tasks

only one source so far

IBM has released Granite 4.2 open-weight models in three sizes (3B, 8B, 30B parameters) with a native context window of 128 000 tokens. The 8B and 30B variants have undergone agentic training for working with the terminal, web search and external tools.

IBM has expanded its family of open-weight language models intended for download and local operation with a new generation, Granite 4.2. The models are available in three sizes – 3B, 8B and 30B parameters – and use a decoder-only architecture. They natively offer a context window of 128 000 tokens.

The 8B and 30B variants have also undergone agentic reinforcement learning training, which, according to IBM, expands their capabilities in areas such as working with the terminal, searching the web or using external tools. The 3B model also offers tool support, but without this specialized training.

According to IBM, this is a release focused on reasoning within the Granite model family, meaning the ability to reason through a chain of intermediate steps (chain-of-thought) and pass intermediate results between individual steps. The source also points out that the term “reasoning” in this context does not mean conscious understanding of a problem as in human reasoning, but the functional ability of a model to break a task down into intermediate steps.

Details about pricing, licensing or hardware requirements are not provided in the source.

What changed

Why it matters

Smaller open-weight models with built-in support for agentic tasks enable local deployment without dependence on cloud APIs, which is relevant to developers building agents with limited budgets or data privacy requirements, as well as companies considering on-premise operation.

Release card

Granite 4.2

IBM

open weights
Specifications
3B, 8B and 30B parameters
Context
128 000 tokens natively
Availability
Available for download and independent local operation (self-hosted)
According to the sources, it is suitable for
  • Working with the terminal (8B and 30B variants)
  • Web search (8B and 30B variants)
  • Using external tools in agentic tasks
Documented limits
  • The 3B variant has not undergone agentic reinforcement learning training like 8B and 30B; it offers tool support without this specialization

The card summarizes information from the article and any dated corrections, with a link to the original source. It is not our assessment of the model. It does not yet have a dedicated editorial profile. Model selection and other announcements →

Two audiences, two different impacts

What this means

01

For individuals

Developers and researchers working with local LLMs are getting smaller open-weight models with built-in support for agentic tasks (terminal, web search, external tools) and a longer context window, without having to rely on a cloud API.

What to do Try deploying Granite 4.2 8B or 30B locally for agentic tasks requiring work with tools or longer context.
More practical updates →
02

For a business

Companies considering local deployment of agentic models (because of costs, data privacy or independence from cloud APIs) are getting an open-weight alternative from IBM in three sizes, with the larger variants incorporating training for working with tools, the terminal and web search.

Development
What to decide Consider testing Granite 4.2 (8B or 30B) for an internal agentic use case requiring local operation and control over data.
More business impacts →
Granite 4.2 IBM LLM open-source reasoning self-hosting

Check the original

Event sources

only one source so far · 1 publisher, 1 independent. We count feeds from the same owner only once.

1
Ars Technica (AI) independent context · first detected IBM's new Granite 4.2 models ride the wave of interest in local LLMs