IBM Granite

Launch of V3 model

IBM Granite is our large language model (LLM) powering our generative AI ecosystem. From understanding and generating natural language conversation, to writing code and digesting large volumes of information, Granite is a general purpose foundation model for business applications.

The LLM market is incredibly crowded, with over 700,000 models on Hugging Face alone. We needed to stand out with an ownable and recognizable visual identity. Additionally, the rapid advancements and change within this landscape necessitated a visual system which was efficient to produce and easy to update.

Our message was simple, Granite is fit-for-purpose. The model is widely capable, comes in a variety of sizes, and integrates a general training corpus with your own proprietary and industry-specific data.

We began with IBM’s core LLM visualization — a nested set of interconnected cubes within a larger package. We used the enclosing shape as a stage upon which we communicated the model’s features, characteristic, and use cases.

The result is an instantly recognizable, flexible communication system.

A library of animations and derivative static visuals are being continuously developed as the model and the market environment change. These assets are distributed widely across media channels and platforms to represent our flagship generative AI platform.

Model characteristics

Input methods

Use cases