DeeperML - #ibm

IBM has announced the release of the Granite 3.2 family of large language models (LLMs), designed to provide efficient AI solutions for enterprises. The new Granite 3.2 models include a vision language model (VLM) that excels in document understanding tasks, rivaling the performance of significantly larger models like Llama 3.211B and Pixtral12B on benchmarks such as DocVQA, ChartQA, AI2D, and OCRBench. IBM also employed its open-source Docling toolkit to process millions of PDFs and generate question-answer pairs, enhancing the VLM's ability to handle document-heavy workflows.

IBM is incorporating conditional reasoning into its Granite 3.2 LLMs, allowing for the optimization of efficiency by enabling users to switch reasoning capabilities on or off. This approach provides flexibility for users to manage intensive processing needs. Additionally, IBM is releasing a new vision model optimized for document processing, aiding in the digitization of legacy documents, and time series forecasting models that apply transformer technology to predict future values from time-based data. All Granite 3.2 models are available under the Apache 2.0 license on Hugging Face, with select models also available on IBM watsonx.ai, Ollama, Replicate, and LM Studio, and expected soon in RHEL AI 1.5.


References :

• AiThority: IBM Expands Granite Model Family with New Multi-Modal and Reasoning AI Built for the Enterprise
• insideAI News: IBM Adds Granite 3.2 LLMs for Multi-Modal AI and Reasoning
• SiliconANGLE: IBM releases new Granite 3.2 family of models that include reasoning when you want it
• AI News | VentureBeat: IBM Granite 3.2 uses conditional reasoning, time series forecasting and document vision to tackle challenging enterprise use cases

Classification:

• HashTags: #IBM #Granite3.2 #LLM
• Company: IBM
• Target: Enterprises
• Product: Granite
• Feature: Granite 3.2
• Type: AI
• Severity: Informative