Two new Large Language Models (LLMs) were released: 

- 𝗚𝗿𝗮𝗻𝗶𝘁𝗲-𝟭𝟯𝗯-𝗰𝗵𝗮𝘁.𝘃𝟮.𝟭: the most advanced version of our Granite models, with superior benchmarks in trust and performance

- 𝗠𝗶𝘅𝘁𝗿𝗮𝗹 𝟴𝘅𝟳𝗯: an open-source model optimized to run on a fraction of computing power

Our clients face increased pressure to transform their business models and they are looking for an evolving library of trusted, high-performing, and cost-effective models.

Here are five things to know about IBM's differentiated approach to foundation models.

𝟭. 𝗖𝗵𝗼𝗶𝗰𝗲𝘀: IBM prioritizes client flexibility in AI, offering a range of IBM-developed, open-source, third-party, and custom model options to best fit diverse business needs and use cases.

𝟮. 𝗧𝗿𝘂𝘀𝘁𝗲𝗱: IBM models, adhering to strict AI ethics and data governance, ensure trust and transparency, backed by intellectual property indemnification.

𝟯. 𝗦𝗽𝗲𝗰𝗶𝗮𝗹𝗶𝘇𝗲𝗱: IBM's enterprise-ready models deliver competitive performance with lower latency and infrastructure requirements, demonstrating significant speed and efficiency advantages in benchmarks.

𝟰. 𝗘𝗳𝗳𝗶𝗰𝗶𝗲𝗻𝘁: our models are designed for reduced GPU use and compute power, offering environmental and cost benefits without compromising on deployment efficiency.

𝟱. 𝗘𝘃𝗲𝗿-𝗲𝘅𝗽𝗮𝗻𝗱𝗶𝗻𝗴: IBM continuously evolves its AI offerings through research innovations and partnerships, expanding capabilities, language support, and introducing new modalities to better meet enterprise needs.

Here is the list of supported foundation models available with watsonx.

Armand