Granite 4.0 1B is a lightweight, open-weights language model developed by IBM as part of its fourth-generation Granite family. Categorized within the Granite 4.0 Nano collection, it is designed for efficiency and speed on resource-constrained devices such as smartphones and IoT hardware. The model is built on a decoder-only dense transformer architecture, featuring Grouped-Query Attention (GQA), RMSNorm, and SwiGLU activation functions.
IBM trained the model on approximately 15 trillion tokens using a multi-stage strategy to ensure performance across general instruction-following and enterprise-specific tasks. Despite its compact size, it supports a context window of 128,000 tokens and natively handles advanced capabilities like tool-calling, retrieval-augmented generation (RAG), and structured JSON output.
Capabilities and Multilingual Support
The model is optimized for high-throughput and low-latency environments, prioritizing instruction-following precision. It supports a diverse range of languages, including English, German, Spanish, French, Japanese, Portuguese, Arabic, Czech, Italian, Korean, Dutch, and Chinese. It is released under the Apache 2.0 license, making it available for both research and commercial use.