IBM Unveils Granite 4.0: Open-Source Language Model for Enterprise Tasks
IBM has unveiled Granite 4.0, the latest iteration of its language models, designed to enhance enterprise tasks and agentive workflows. This new version combines Mamba and Transformer layers in a 9:1 ratio, significantly reducing RAM usage.
Granite 4.0 is built on a hybrid architecture that merges Mamba-2 and Transformer layers, with nine parts Mamba and one part Transformer. This design allows for up to 70 percent less RAM usage compared to pure Transformer models, as reported by IBM. The models are released open-source under Apache 2.0, cryptographically signed, and accredited to ISO/IEC 42001:2023.
IBM positions Granite 4.0 as modular components for agentive workflows and standalone models for enterprise tasks. The service provider for integrating Granite 4.0 into tools like Unsloth and Continue is IBM Consulting. They offer these models and support integration and automation solutions for enterprise workflows.
Granite 4.0 is trained on a 22T-token corpus compiled from various enterprise-relevant sources. It includes four model variants: Granite-4.0-H-Small, Granite-4.0-H-Tiny, Granite-4.0-H-Micro, and Granite-4.0-Micro. IBM grants an unlimited waiver for third-party IP claims for content generated by Granite on IBM watsonx.ai, ensuring users can safely utilise these models.
IBM's Granite 4.0, with its innovative hybrid architecture and enterprise-appropriate training data, promises to streamline tasks and workflows in business environments. Its open-source availability and integration support from IBM Consulting make it an attractive option for enterprises looking to leverage advanced language models.
Read also:
- Worldcoin's Iris Scanning Raises Global Privacy Concerns
- Exploring Harry Potter's Lineage: Decoding the Enigma of His Half-Blood Ancestry
- Elon Musk Acquires 26,400 Megawatt Gas Turbines for Powering His AI Project, Overlooks Necessary Permits for Operation!
- Machine Learning and Forecasting Advancements in Supply Chain Resilience: Insights from Jaymalya Deb
