On October 21st local time, IBM released the latest version of its enterprise oriented artificial intelligence model, Granite 3.0, which adopts an open source model. This business model is different from competitors such as Microsoft, Amazon Web Services, etc., which charge users to access artificial intelligence models.
Although IBM's artificial intelligence model adopts an open source model, it is available on IBM's software platform WatsonX, which adopts a paid model. Through the WatsonX tool, enterprises can use customized AI models running within data centers.
Regarding the openness of Granite 3.0 in China, IBM responded to First Financial reporters, saying, "All models of Granite 3.0 are open source, and developers can obtain it for free through open source communities such as GitHub and Hugging Face. When using it, they follow the general practices of open source software according to the standard Apache 2.0 license requirements. As long as Chinese developers can access the above open source communities normally, they can also obtain and use Granite 3.0 normally
The latest release of Granite 3.0 uses Nvidia H100 GPU for training, and IBM will also provide some large model versions on Nvidia's software tool stack.
Last month, IBM just announced the launch of an instance on its cloud platform equipped with Nvidia H100 GPU. Users can use Nvidia's GPUs for artificial intelligence workloads, including training and inference. In addition, IBM Cloud Platform also offers NVIDIA L40S and L4 GPU, which can be used for smaller AI workloads such as training small models or deploying chatbots.
IBM has been striving to reshape its leadership position in the industry through hybrid cloud and AI in recent years. But at the same time, IBM's cloud business is also undergoing a new round of adjustments, and the company expects artificial intelligence to replace nearly 8000 jobs. In August of this year, IBM announced the closure of its research and development department in China, involving over 1000 employees.