Skip to content

IBM Granite 3.1

Enterprise models with a longer context window

Visit official site

What IBM Granite 3.1 is

The series is described as an enterprise-focused language model family covering tool use, retrieval augmented generation and scalable agentic workflows. It expands the context window to 128K tokens and adds a retrieval-optimised embedding family.

What you can do with it

  • Pick a model such as Granite 3.1 8B Instruct
  • Fetch models through watsonx.ai or partner platforms
  • Configure the model for a specific enterprise use case

Application scenarios

  • Build document understanding and question answering
  • Power semantic and vector search with embeddings
  • Add control and observability to tool calls

Pros & cons

✓ What we like

  • Function-call hallucination detection is added
  • Models are available through watsonx.ai, Hugging Face and Docker

! What to watch out for

  • Multimodal capabilities are planned for the first quarter of 2025
  • Model selection depends on the chosen enterprise use case

FAQ

How long is the context?

The material states a 128K token context window for the Granite 3.1 models.

Under which licence is it released?

All Granite 3.1 models are released under Apache 2.0.

Last reviewed: 2026-11-15

More AI coding tools tools

View all →

How we review