Granite 3.3 8b Instruct is an open-weights instruction-tuned language model hosted on Hugging Face under the ibm-granite organization. It forms one implementation within the class of foundation models and is supplied with a chat template that defines its system prompt and tool-calling behavior.
The supplied chat template sets a default system message that includes a knowledge cutoff date of April 2024, the current date, and the statement that the model is developed by IBM. When tools and documents are provided, the template instructs the model to respond only with a specific tool-call token followed by a JSON list of tools when a tool is required, to notify the user if a requested tool is unavailable, and to align responses strictly with facts present in supplied documents while informing the user when needed information is absent. The template also contains logic to alias an input field named tools to available_tools if the latter is not supplied.
The model is delivered as a downloadable repository on the Hugging Face platform, where users can access its weights, configuration, and the associated chat template. No pricing, licensing terms, or additional deployment methods are stated in the repository page itself.
Granite 3.3 8b Instruct is a Foundation models & chat product. It focuses on running high-quality open instruction-tuned LLMs locally or in the cloud without proprietary vendor lock-in. It is built as an open-source project for developers. Granite 3.3 8b Instruct is open source under the Apache-2.0 license. It runs on the web and the command line, and it can be self-hosted.
IBM builds and maintains Granite 3.3 8b Instruct, and the product first shipped in 2024. Development happens publicly on GitHub with 388 stars. Key capabilities include Instruction Tuning, Tool Calling, and Document Grounding.
Latest indexed changes and source events
ibm-granite/granite-3.3-8b-instruct verified by the PulseGate indexer
