IBM announced a set of enhancements to its watsonx platform that broaden generative Artificial Intelligence capabilities and expand support for third-party foundation models. The company introduced Granite, a new series of models built on a decoder architecture intended to improve sequence generation for enterprise natural language processing tasks. IBM said Granite will arrive on watsonx in Q3 2023. In addition to Granite, watsonx has been extended to support the fine-tuned chat version of Meta’s Llama 2 and StarCoder, the code generation model from Hugging Face and ServiceNow; those third-party models are already accessible to watsonx users.
Beyond new models, IBM is adding tooling to help enterprises customise and validate deployments. A Tuning Studio will let users refine prompts and adapt models to business needs, with planned availability in Q3 2023. The platform also includes a synthetic data generator to create artificial datasets for testing models, which IBM lists as available now. To address trust and transparency, IBM announced a tech preview of the watsonx.governance toolkit, positioned to support explainable Artificial Intelligence and governance workflows; the toolkit was previously teased in July.
The updates are part of a wider roll‑out strategy for the watsonx platform. IBM began enterprise rollout in July and plans to add further models and tools over time to support production use cases. The company framed the changes around meeting enterprise requirements for customization, testing and oversight. “As demonstrated by the ongoing rollout of the watsonx platform within just a few months since launch, we are here to support clients through the entire Artificial Intelligence lifecycle,” said Dinesh Nirmal, president of products at IBM Software, in a statement.
