FrontierAugust 26, 2026via MarkTechPost
IBM Releases Granite 4.2: Bringing Native Reasoning and Agentic RL to Open Enterprise Models
Why it matters
IBM ships open-weight reasoning models with native agentic capabilities (code editing, terminal control, web search in sandbox) across three sizes. Granite 4.2 expands the frontier lab race into the open-source tier with reasoning-on-demand and production-ready agent training — a direct challenge to the closed-model moat on reasoning and autonomy.
Key signals
- Three model sizes: 3B, 8B, 30B
- Apache 2.0 license (fully open)
- Thinking/low-effort/non-thinking mode switcher on all sizes
- Native tool calling across all sizes
- Agentic RL training on 8B and 30B (code editing, terminal control, web search in sandbox)
- 30B benchmarks: 57.00 SWE-Bench Verified, 29.24 Terminal-Bench 2.1
- Open-source reasoning competes with proprietary frontier models
The hook
IBM's Granite 4.2 brings reasoning and agentic RL to open weights — 30B hits 57 on SWE-Bench, no proprietary license.
IBM has released Granite 4.2, a family of open reasoning language models in 3B, 8B, and 30B sizes, all under Apache 2.0. Every model exposes a thinking / low-effort / non-thinking switch and native tool calling. The 8B and 30B additionally go through an agentic RL block that trains them to edit code…