FrontierThe story, in brief
IBM drops open-weight Granite 4.2 family with built-in agentic capabilities under Apache 2.0
IBM drops Granite 4.2 open-weight models with native agentic RL — 512K context, Apache 2.0, no proprietary tax.
The KeyNews take
Why it matters
IBM's open-weight Granite 4.2 family introduces agentic reinforcement learning capabilities directly in the base models at accessible sizes (3B-30B), lowering the barrier for practitioners to deploy autonomous tool-use behavior without proprietary model licensing. The 512K context window and Apache 2.0 terms make this a notable entry in the open-weight capability race.
The key facts
6 to knowThree model sizes: 3B, 8B, 30B
15 trillion tokens training data
512,000 token context window
Agentic RL training for tool use and code execution
Apache 2.0 license (permissive open-weight)
Released by IBM
Go to the source
The Decoderthe-decoder.com
Publisher excerpt: IBM is releasing its Granite 4.2 language models in 3B, 8B, and 30B sizes, trained on about 15 trillion tokens with a context window of up to 512,000 tokens. The larger models use "agentic RL" training to learn tool use and code execution on their own. All models are available under the Apache 2.0…