IBM Launches Granite 4.1 Model Family With Vision, Speech, and Safety AI Capabilities for Enterprise Use

IBM launches Granite 4.1, its most expansive AI model family yet, featuring small language, vision, speech, embedding, and safety models built for enterprise use, all released under an Apache 2.0 license with state-of-the-art performance across document understanding, multilingual transcription, and harm detection.

• IBM releases the Granite 4.1 model family, its most expansive to date, covering small language models in 3B, 8B, and 30B sizes, along with vision, speech, embedding, and guardian models designed for enterprise AI workflows.
• Granite 4.1 language models deliver competitive instruction-following and tool-calling performance without reasoning chains, trained on 15 trillion tokens with a multi-stage reinforcement learning pipeline, while Granite Vision 4.1 and Speech 4.1 achieve state-of-the-art results in document understanding and multilingual transcription respectively.
• All Granite 4.1 models are released under an Apache 2.0 license and are optimized for popular inference runtimes including vLLM, SGLang, and llama.cpp, with Granite Guardian 4.1 providing expanded harm detection and safety moderation capabilities across any AI pipeline.