Granite 4.1 3b is a 3-billion-parameter language model released by IBM in GGUF format on Hugging Face. The repository provides the quantized model files along with system prompt templates that define its behavior for specific tasks.
The included templates configure the model to support tool calling through a defined XML-based format. When tools are supplied, the model receives instructions to return function calls inside designated XML tags containing JSON objects with a function name and arguments. If a requested tool is absent from the provided list, the model is directed to inform the user it cannot fulfill the request. A separate template addresses document-grounded responses. It instructs the model that it has access to a list of documents enclosed in XML tags and must align all answers strictly with facts from those documents. When the required information is unavailable, the model must state that the question cannot be answered based on the available data.
The model is delivered as GGUF files suitable for local inference. The repository forms part of IBM's open Granite model family and is hosted on the Hugging Face platform, which supports open-source and open-science initiatives. No pricing, licensing terms, or additional deployment details are specified in the repository page.
Granite 4.1 3b sits in PulseGate's Foundation models & chat category. It focuses on running efficient, locally-hosted language model inference with tool-calling capabilities on consumer hardware. It is built as an open-source project for developers and AI practitioners. Granite 4.1 3b is open source under the Open Source license. The product ships for the web, the command line, and API.
IBM builds and maintains Granite 4.1 3b, and the product first shipped in 2024. It operates in a well-populated space: PulseGate tracks 7 similar tools.
Latest indexed changes and source events
ibm-granite/granite-4.1-3b-GGUF verified by the PulseGate indexer
