AI provider / IBM
IBM
A provider linked to enterprise-first open models, trusted AI positioning and practical deployments across business systems.
IBM brings AI to the enterprise on its own terms: governed, trusted and production-ready. The Granite family, delivered through watsonx, is built as small, open and efficient workhorses tuned for real business workloads rather than leaderboard headlines.
From the first Granite 13B to the Mamba-2/transformer hybrid of Granite 4.0 and the speech-and-vision breadth of 4.1, the line emphasizes governance, enterprise integration and Apache-licensed openness — including a dedicated Granite Code branch for software delivery.
Model catalog
Every release in this provider’s model line.
Each row is a distinct model release — its date, its type and the traits it is known for. Open a source to dig into any of them.
| Model | Released | Type | Key traits | Source |
|---|---|---|---|---|
| Granite 4.1 | Apr 2026 | Model family |
| IBM Research |
| Granite 4.0 | Oct 2025 | Architecture |
| IBM Granite docs |
| Granite 3.2 | Feb 2025 | Version update |
| IBM Research |
| Granite 3.1 | Dec 2024 | Version update |
| IBM Granite docs |
| Granite 3.0 | Oct 2024 | Open release |
| IBM |
| Granite Code | May 2024 | Coding |
| GitHub |
| Granite 13B | Nov 2023 | Debut |
| Wikipedia |
Timeline
A practical timeline of model releases and company milestones.
This page tracks the moments that matter most: major model releases, version jumps and company moves that changed how this provider shows up in the AI market.
Granite 4.1 rounds out the family across speech, vision and safety.
Following April’s 4.0 3B Vision for document extraction, the 4.1 collection spans language, speech, vision, embeddings and Guardian.
Granite 4.0 debuts a hybrid Mamba-2/transformer architecture.
70% lower memory and 2x faster inference make it one of the first production bets on post-transformer designs.
Granite 3.2 adds toggleable reasoning and vision.
Chain-of-thought you can switch off to save compute, plus a document-vision model; 3.3 follows in April with speech.
Granite 3.1 stretches context to 128K across the family.
A fast follow-up adds long context and improved RAG performance, holding a quarterly release cadence.
Granite 3.0 resets the line as small, open enterprise workhorses.
The 2B and 8B generation embraces a “small but fit-for-purpose” thesis, with Guardian safety models alongside.
Granite Code models go open source under Apache 2.0.
IBM open-sources a full code-model family from 3B to 34B, pivoting Granite from closed enterprise tool to open ecosystem.
Granite 13B opens IBM’s enterprise model line on watsonx.
Chat and instruct variants trained on curated business data, sold with IP indemnification — enterprise trust as the product.