- New Granite 3.0 8B and 2B models, released under a fully permissive Apache 2.0 license, deliver state-of-the-art performance in academic and enterprise benchmarks, outperforming or matching similarly sized models
- New Granite Guardian 3.0 models offer more comprehensive protection with features for safe and reliable AI
- IBM Unveils Next-Generation Watsonx Code Assistant, Powered by Granite, for General-Purpose Coding; Also Launching Tools on Watson.ai for Building and Deploying AI Agents and Applications
IBM (NYSE: IBM) announced during its TechXchange conference, the launch of the family of AI models, the Granite 3.0. IBM’s third-generation Granite language models can outperform or match similarly sized models from leading vendors on many academic and industry benchmarks, demonstrating solid performance, transparency and security, the company says.
In keeping with IBM’s commitment to open source AI, Granite models are released under a fully permissive Apache 2.0 license, making them unique in the combination of performance, flexibility and rights they offer to enterprise customers and the broader community.
IBM's Granite 3.0 family includes:
• General Use/Language: Granite 3.0 8B-Instruct, Granite 3.0 2B-Instruct, Granite 3.0 8B Base, Granite 3.0 2B Base
• Security: Granite Guardian 3.0 8B, Granite Guardian 3.0 2B
• Combination of experts: Granite 3.0 3B A800M Instruct, Granite 3.0 1B A400M Instruct, Granite 3.0 3B A800M Base, Granite 3.0 1B A400M Base
The new Granite 3.0, 8B, and 2B language models are designed to support large workloads and facilitate the adoption of enterprise AI, delivering excellent performance and cost-effectiveness for tasks such as augmented regeneration (RAG), classification, summarization, entity extraction, and tooling. These compact and versatile models are designed to accurately fit business data and integrate seamlessly into any business environment or workflow.
The launch of Granite 3.0 reaffirms IBM's commitment to transparency, security and trust.
IBM provides intellectual property (IP) indemnification for all Granite models on watsonx.ai, enabling enterprise clients to have greater confidence in combining their data with these models.
Raising the Bar: Granite 8B and 2B Benchmarks
On basic RAG business tasks using cybersecurity tools and tasks, the Granite 3.0 8B Instruct model shows superior average performance compared to similarly sized open source models from Mistral and Meta[1].
The models were trained on over 12 trillion tokens on data drawn from 12 different natural languages and 116 different programming languages using a novel two-phase training method leveraging the results of several thousand experiments designed to optimize data quality, data selection, and training parameters. By the end of the year, the 8B and 2B models will also include support for extended context length of 128K and multimodal document understanding capabilities.
Introducing Granite Guardian 3.0: Ushering in the Next Era of Responsible AI
As part of this release, IBM is also introducing a new family of Granite Guardian models that enable application developers to implement security safeguards by checking user prompts and LLM responses for a variety of risks. The Granite Guardian 8B and 2B models provide the most comprehensive set of risk and damage detection capabilities available on the market today.
In addition to harm dimensions such as social bias, hate, toxicity, profanity, violence, jailbreaking, and more, these models also provide a variety of unique RAG-specific checks such as substantiation, context relevance, and response relevance.
While Granite Guardian models are derived from the corresponding Granite language models, they can be used to implement security measures alongside any AI model, whether open or proprietary.
Granite 3.0 model availability
The entire Granite 3.0 model suite and updated time series models are available for download from Hugging Face under the permissive Apache 2.0 license. Instructional variants of the new Granite 3.0 language models, 8B and 2B, and the Granite Guardian 3.0 models will be available for commercial use on IBM’s WatsonX platform. A selection of Granite 3.0 models will be available as NVIDIA NIM microservices and through Google Cloud’s Vertex AI Model Garden integrations with HuggingFace.
From Assistants to Agents: Making the Future of Enterprise AI a Reality
IBM is advancing enterprise AI across a spectrum of technologies – from models and assistants to the tools needed to tune and deploy AI specifically for enterprises’ unique data and use cases. IBM is also paving the way for future AI agents that can self-direct, reflect and perform complex tasks in dynamic business environments.
Additionally, IBM today introduced the next version of watsonx Code Assistant, based on Granite code templates, to provide general-purpose programming assistance in languages such as C, C++, Go, Java, and Python, with advanced application modernization capabilities for enterprise Java applications. Granite code capabilities can now also be accessed through a Visual Studio Code extension, IBM Granite.Code.
IBM will continue to develop agent capabilities across its portfolio in 2025, including pre-built agents for specific domains and use cases.
Extending the AI Platform to Empower IBM Consultants with AI
IBM is also announcing a major expansion of its AI-powered delivery platform, IBM Consulting Advantage. The multi-model platform contains AI agents, applications, and methods, as repeatable frameworks, that empower 160,000 IBM consultants to deliver improved, faster client value at lower cost.
Another key part of the expansion is the introduction of IBM Consulting Advantage for Cloud Transformation and Management and IBM Consulting Advantage for Business Operations. Each includes domain-specific AI agents, applications and methods infused with IBM IP and best practices so consultants can accelerate clients’ cloud and AI transformations across tasks such as code modernization, quality engineering or transforming and running operations across domains such as finance, HR and procurement.
For more information about this announcement, Click here.
To learn more about Granite and IBM’s AI strategy for business, visit this link.
EN 
PT 
