IBM has officially launched its third generation of Granite large language models (LLMs), known as Granite 3.0, aimed at enhancing enterprise artificial intelligence (AI). This move comes as IBM seeks to expand its already substantial $2 billion business in generative AI. The new models are designed to support various enterprise applications, including customer service, IT automation, business process outsourcing, application development, and cybersecurity.
New Features of Granite 3.0
The Granite 3.0 models include both general-purpose options with 2 billion and 8 billion parameters, as well as specialized Mixture-of-Experts (MoE) models. These MoE models consist of variants such as Granite 3.0 3B A800M Instruct and Granite 3.0 1B A400M Base. Additionally, IBM has introduced models with enhanced safety features, branded as Granite Guardian 3.0, which aim to prevent misuse and ensure the responsible deployment of AI technology.
Rob Thomas, IBM’s Senior Vice President and Chief Commercial Officer, highlighted the rapid growth of the company’s generative AI business, stating, “As I think about my 25 years in IBM, I’m not sure we’ve ever had a business that has scaled at this pace.”
Training and Performance
The Granite 3.0 models were developed using a training process that involved 12 trillion tokens of data, which included both language and code data. Dario Gil, IBM’s Senior Vice President and Director of Research, emphasized that the quality of the data and the architectural innovations used in training set these models apart from previous generations.
According to Gil, the Granite models have achieved impressive results, outperforming competitors like Google and Anthropic in various tasks. He stated, “What you’re seeing here is incredibly highly performant models, absolutely state of the art, and we’re very proud of that.”
In addition IBM states it has prioritized safety in its AI models. The newly introduced Guardian models are designed to prevent core models from being manipulated or generating harmful content. Gil noted the importance of balancing performance with inference costs, which are crucial for scaling AI applications in enterprises.
Open Source Commitment
A significant aspect of the Granite 3.0 launch is IBM’s commitment to open-source technology. The models are released under the Apache 2.0 open-source license, which allows enterprises to customize and build upon the technology without restrictions. This approach contrasts with competitors like Microsoft, which charge for access to their models.
Gil explained the rationale behind this decision: “We decided that we’re going to be absolutely squeaky clean on that, and decided to do an Apache 2 license, so that we give maximum flexibility to our enterprise partners.” This flexibility is expected to foster a vibrant ecosystem of applications and solutions powered by Granite technology.
Thomas added that this open-source model could significantly accelerate the adoption of AI in businesses. “It’s completely changing the notion of how quickly businesses can adopt AI when you have a permissive license that enables contribution, enables community and ultimately, enables wide distribution,” he said.
Future Directions: Generative Computing
Looking ahead, IBM is exploring the concept of generative computing, which refers to programming computers through examples or prompts rather than traditional step-by-step instructions. This innovative approach aligns with the capabilities of LLMs like Granite, which can generate text, code, and other outputs based on user input.
Gil described this shift as fundamental, stating, “You are going to see us invest and go very aggressively in a direction where with this paradigm of generative computing, we’re going to be able to implement the next generation of models, agentic frameworks and much more than that.” This vision reflects IBM’s commitment to leading the next wave of AI development.
Availability and Integration
Some of the new Granite models are now available for commercial use on IBM’s Watsonx platform, which allows businesses to run and customize the models within their own data centers. Some models will also be accessible through Nvidia’s software tools, which are designed to allow businesses to integrate AI technologies.
The aggressive foray of IBM into the AI model space aims to meet the growing needs of enterprises looking to leverage AI for various applications. With a focus on open-source principles, safety, and performance, IBM aims to solidify its place as a leader in the enterprise AI market.
(Photo by Carson Masterson)
