IBM worries general purpose LLMs are a black boxes and so are unsuited for enterprises especially in the regulated sectors. To tackle this problem, they trained an LLM of their own Granite which is an open source model under Apache 2.0 license for anyone to use for their own commercial purposes. To train this model they filtered out hate speech, abuse, profanity, and removed any datasets at risk of copyright breach or pirating. The idea is to have a specialized model suited for enterprises instead of a general-purpose model trained on generic data 1/5th of the data the model is trained on is from finance and legal domains making it highly relevant for enterprises in those industries
They also have watsonx an AI and data platform that enables enterprises to use their own data to train, validate, tune and deploy AI systems across the business And then they have InstructLab which can be used by enterprises to customize Granite 7B model using their own data
There are several benefits of open source models are cost efficiency, flexibility and transparency. Compare these to the benefits of closed source models are higher performance, easier implementation, and streamlined security and privacy controls.
References: Reuters Plus. (2024). IBM: From black box to open book. https://plus.reuters.com/ibm-from-black-box-to-open-book/p/1 ODSC - Open Data Science. (2025, Jan 10). The benefits of open-source vs. closed-source LLMs. Medium. https://medium.com/@ODSC/the-benefits-of-open-source-vs-closed-source-llms-67990c000787