IBM unveiled the latest iteration of its artificial intelligence models, Granite 3.0, on Monday, aiming to capitalize on the growing adoption of generative AI in enterprises. Unlike some of its competitors, such as Microsoft (MSFT.O), which charge for access to AI models, IBM is making its Granite models open-source, continuing the approach seen with other versions in the Granite family.
IBM also offers Watsonx, a paid tool that allows businesses to customize and run AI models within their own data centers. Starting Monday, select variants of the new Granite models will be commercially available on the Watsonx platform, providing businesses with more flexibility in deploying AI. In addition, some of these models will be accessible through Nvidia’s software stack, enabling companies to integrate AI into their operations more seamlessly.
According to Dario Gil, IBM’s director of research, the Granite models were trained using Nvidia’s H100 GPUs, one of the leading chips in AI computing. This partnership underscores the role of high-performance computing in driving advancements in enterprise AI solutions