Granite
An open-source AI model family for on-premises deployment, licensed under Apache 2.0, with language, speech, vision and guardrail models.
Granite is IBM's family of open-source AI models for developers and businesses that want to run and customize AI on their own hardware or servers. The language-model repository listed here is archived and no longer maintained. The broader family includes models for language, speech, document understanding and forecasting, released under Apache 2.0 for research and commercial use.
Local deployment is supported through Ollama and LM Studio, with models available on Hugging Face. Hosted access is also available through services including watsonx.ai, OpenRouter and Replicate. Those services run models in the cloud; downloaded models can run on your own infrastructure.
The language models handle reasoning, coding and multi-step tool use for AI agents. Developers can also use them for document summarization and answers grounded in an external knowledge base. Granite Vision interprets documents, charts and images, while Speech handles transcription across accents and noisy recordings. Embedding models support semantic search and classification. The time series models provide forecasting, including compact models that can run on a laptop.
Granite's emphasis is on enterprise customization and smaller models with lower compute demands. IBM publishes information about training data and processes, and its data review covers licensing and governance. Granite Guardian detects harmful prompts and model responses, as well as malicious content. IBM cryptographically signs its released language, vision, speech, embedding and guardian models.
