Skip to content

IBM Granite 3.0 models

Run IBM language models through Ollama

Visit official site

What IBM Granite 3.0 models is

The models are described as a family of AI language models from IBM served on the Ollama platform. Training used more than 12 trillion tokens, and both dense and MoE variants exist.

What you can do with it

  • Summarise, classify and extract text
  • Run RAG and question answering tasks
  • Generate, translate or fix code

Model notes

  • Dense and MoE variants are listed
  • MoE variants target low-latency on-device use
  • Versions named include 2B, 8B and MoE

Pros & cons

✓ What we like

  • Summarisation, classification and extraction are listed
  • MoE variants are aimed at on-device and instant inference

! What to watch out for

  • Model versions include 2B, 8B and MoE choices
  • Access goes through the Ollama platform

FAQ

Who developed them?

The material states the models were developed by IBM and provided through Ollama.

What tasks are listed?

Summarisation, classification, extraction, question answering, RAG and code work.

Last reviewed: 2026-11-15

More AI coding tools tools

View all →

How we review