Granite is IBM's family of open AI models for businesses, released under the Apache 2.0 licence, covering language, code, vision, speech, embeddings, safety and time series.
In plain words
IBM builds Granite for companies that want models they can inspect, customise and run on their own servers. The models are kept fairly small so they are cheaper to run. Besides chat and reasoning models, the family includes Guardian models that flag harmful content, embedding models for search, and time-series models for forecasting. Weights are shared on Hugging Face.
A quick example
Imagine a bank that must keep customer data in its own data centre. It could download a Granite language model and a Granite Guardian model, run both on-premises, and have the Guardian model screen the answers before staff see them.
Where you'll meet it
- On-premises enterprise AI and document understanding.
- Guardrails, retrieval and time-series forecasting.
Related
Sources
- DocsIBM Granite, IBM · read 29 Sept 2026
- DocsIBM Granite, IBM (GitHub) · read 29 Sept 2026
Checked by WhatIsAI against the sources above · 29 Sept 2026. Spotted an error? Report it.