IBM has announced the release of its Granite 3.0 open-source AI models for enterprise clients as the tech giant is looking to strengthen its market position and capitalize on the increased adoption of generative AI (GenAI) technology by businesses.
The company unveiled the new AI model at its annual TechXchange event. It said the third-generation Granite flagship language models can match and even outperform other similarly sized models from other providers.
The new Granite 3.0 8B and 2B language models were designed to perform in the workplace setting, which was why the company also dubbed it as the “workhorse” model for the enterprise AI, according to the company.
IBM noted that the models can handle complex processes such as “Retrieval Augmented Generation (RAG), classification, summarization, entity extraction, and tool use.”
They were also “designed to be fine-tuned with enterprise data and seamlessly integrated across diverse business environments or workflows.”
IBM also shared the result of Granite 3.0’s raw performance using some benchmarks. Using HuggingFace’s OpenLLM Leaderboard, the Granite 3.08B Instruct model leads on average when compared to similar-sized open-source models from Meta and Mistral. In terms of safety, IBM used the AttaQ safety benchmark.
“The Granite 3.0 models were trained on over 12 trillion tokens on data taken from 12 different natural languages and 116 different programming languages, using a novel two-stage training method, leveraging results from several thousand experiments designed to optimize data quality, data selection, and training parameters,” the company stated.
Reuters reported that the “Granite 3.0” will differ with other rivals like Microsoft’s, where customers had to pay to access their models. IBM is offering Watsonx, a paid tool that helps in running models from within data centers after they have already been customized.
IBM also updated its pre-trained Granite Time Series models, the first versions of which were released earlier this year. The company claims these new models have outperformed 10 times larger models from Google, Alibaba, and others.
IBM also unveiled its next generation of Granite-powered watsonx Code Assistant for general purpose coding. It also announced that Granite will become the default model of Consulting Advantage, an AI-powered delivery platform used by IBM’s 160,000 consultants.
Some of the variants of the new Granite models will be made available starting October 21 for commercial use on IBM’s watsonx platform. A number of the new models will also be on NVIDIA and through Google Cloud’s Vertex AI Model Garden integrations with HuggingFace.
IBM’s latest AI model comes as more businesses are adopting Gen AI to improve efficiency and analyze and interpret huge volumes of data. IBM noted that GenAI is now more than $2 billion book of business plus across technology and consulting, according to VentureBeat
