IBM launches Granite 4.2 models with enhanced local LLM capabilities

IBM has introduced the latest iteration of its Granite series, a family of open-weight large language models (LLMs) designed for download and self-hosting. The Granite 4.2 models are available in three parameter variants: 3B, 8B, and 30B. These models continue to use a decoder-only architecture, a design choice consistent with previous versions. All variants feature a native context window of 128,000 tokens, allowing for extensive text processing. The 8B and 30B models include an agentic reinforcement-learning block, which enhances their ability to perform tasks such as terminal usage, web searching, and integration with external tools. The 3B model also supports tool integration, though it lacks the specialized training that the larger variants have received. IBM's release underscores the growing interest in local LLMs, which prioritize data privacy and reduced dependency on cloud infrastructure. The company's approach aligns with industry trends toward decentralized AI solutions, offering organizations greater control over their language model operations. The Granite 4.2 series represents a significant step forward in IBM's efforts to provide flexible, powerful, and secure AI tools for a wide range of applications.