At IBM’s annual TechXchange event, the company announced its most advanced AI model to date, Granite 3.0. IBM’s third-generation flagship Granite language models outperform or equal similar-sized models from leading model vendors across multiple academic and industry benchmarks, demonstrating robust performance, transparency, and safety.
IBM unveils its most advanced AI model to date, Granite 3.0.
In line with the company's commitment to open source AI, Granite models are released under the permissive Apache 2.0 license, making them unique in the combination of performance, flexibility, and autonomy they offer to enterprise customers and the community at large.
The new Granite 3.0 8B and 2B language models are designed as “go-to” models for enterprise AI, delivering strong performance for tasks such as Retrieval Augmented Generation (RAG), classification, summarization, entity extraction, and tool usage. These compact, flexible models are designed to be fine-tuned with enterprise data and seamlessly integrated across a variety of business environments or workflows.
While many large language models (LLMs) are trained on publicly available data, the vast majority of enterprise data remains untapped. By combining a small Granite model with enterprise data, specifically using the revolutionary alignment technique InstructLab — introduced by IBM and RedHat in May — IBM believes that enterprises can achieve task-specific performance on par with larger models at a fraction of the cost (based on a range of 3x-23x lower costs than large frontier models in some early proofs of concept).
Granite 3.0 AI Model.
The Granite 3.0 release reaffirms IBM’s commitment to building transparency, safety, and trust in AI products. The Granite 3.0 technical report and responsible use guide provide descriptions of the datasets used to train these models, details on the filtering, cleaning, and curation steps applied, and comprehensive results on model performance across key academic and enterprise benchmarks.
The “Granite 3.0” models are open source, similar to other versions of IBM’s Granite AI family. This approach differs from competitors like Microsoft, which charge customers for access to the models. In return, IBM offers a paid tool called Watsonx, which runs the models inside data centers after they’ve been customized.
Several variants of the new Granite models are available for commercial use starting today on the Watsonx platform. Some of these models will also be available on Nvidia’s software toolkit, allowing businesses to incorporate AI models.
The new Granite models were trained using H100 graphics processing units (GPUs) from Nvidia, a leader in AI chips, said Dario Gil, director of research at IBM.
(According to IBM, Reuters)
Source: https://www.baogiaothong.vn/ibm-phat-hanh-mo-hinh-ai-moi-cho-doanh-nghiep-uu-viet-hon-microsoft-192241021144242992.htm
Comment (0)