JAKARTA - IBM launched Granite 3.0, the latest artificial intelligence (AI) model line designed for high-performance business needs, offers transparency, and improves security.
IBM claims that this model has an equal or better performance with other models, but with the 3 to 23 times lower than the large frontier model.
This launch marks the beginning of the era of a small, high-performance AI model that when integrated with company data, is able to provide optimal performance to perform specific tasks that are equivalent to a larger model, but at a much lower cost," the company said in its official statement quoted on Sunday, November 17.
Granite 3.0 is trained using 12 trillion tokens in 12 languages so that it can be adjusted to the company's unique data, and is easily integrated into various scenarios or workflows of the company.
Because it offers transparency, when used, this new IBM AI model will reveal details of training datasets, including filtration measures and data curation.
Even better, this Granite model is even able to detect risks such as bias, toxicity, and violence, as well as perform unique specific checks such as connectedness, context relevance, and relevance of answers.
SEE ALSO:
This latest model includes several sizes and types, such as:
- Granite 3.0 Instruct & Base for general purposes.- Granite Guardian for security.- Granite Mixture-of-Experts for efficient inference.- Granite 3.0 is available on various platforms, including IBM watsonx, Nvidia, Google Vertex, and ecosystem partners such as AWS and SAP.
The English, Chinese, Japanese, Arabic, and French versions are automatically generated by the AI. So there may still be inaccuracies in translating, please always see Indonesian as our main language. (system supported by DigitalSiber.id)