IBM has launched four new artificial intelligence models that are small enough to run on a standard laptop or even within a web browser. Carl Franzen reports for VentureBeat that these Granite 4.0 Nano models represent a shift in AI development, prioritizing efficiency over massive size. The models are open source and available under the Apache 2.0 license, which permits free use for research, development, and commercial applications.
The new models range from 350 million to 1.5 billion parameters, a small fraction of the size of models from companies like OpenAI or Google. Despite their compact size, IBM claims they are highly capable. According to benchmark results shared by the company, the Granite Nano models outperform competing models in their class on specific tasks such as instruction following and tool usage.
This focus on small, local models offers significant advantages. It allows developers to build applications for devices with limited computing power and enhances user privacy, as data can be processed on the device itself without being sent to the cloud. By making the models open and accessible, IBM aims to provide an alternative to the large, proprietary AI systems that currently dominate the market.
The company is actively engaging with the open-source community to support developers and has indicated that larger and more advanced models are already in development.
