IBM Granite 3.0 models, available via Ollama, include both dense and Mixture of Experts (MoE) versions, each licensed under Apache 2.0. The Granite dense models, Granite 2B and Granite 8B, are text-only LLMs trained on over 12 trillion tokens, showing improved performance and speed, with Granite 8B Instruct rivaling Llama 3.1 8B Instruct on OpenLLM Leaderboard benchmarks. These models are suited for tool-based applications and support retrieval augmented generation (RAG), facilitating tasks like code generation and translation. Meanwhile, the 1B and 3B Granite MoE models cater to low-latency, on-device applications, trained on over 10 trillion tokens for instantaneous inference. These models offer capabilities such as summarization, text classification, text extraction, question-answering, and multilingual dialog use cases.