Anil K Nayak’s Post

Today IBM released IBM Granite 3.0, a family of task targeted large language models. I’m particularly interested in Granite Guardian a model trained for monitoring the input and output of a ‘core’ model to shield it from prompt attacks and other threats such as inappropriate output and jail breaking. When applied to RAG use cases it can also be used to validate answer relevance and groundedness; reducing the threat of hallucinations.

Today, IBM announced Granite 3.0. These models are open, fit-for-purpose and deliver high performance across various industry benchmarks, outperforming or on par with leading industry models! Granite 3.0 features include: Agentic support, RAG and Summarization, Entity Extraction and classification. Read the announcement -> https://lnkd.in/esN9PccD Check out Granite 3.0 -> https://ibm.biz/Bdav8H #AI #GenAI #IBMTechXchange2024

💡What a way to start the week, this morning at TechXchange IBM launched Granite 3.0, a game-changing technology that brings the power of AI to IBMers. You can learn more about today’s release here:  https://lnkd.in/ddWxiQwp #AI #IBM #technology #opensource #innovation

Meet IBM Granite 3.0, advanced AI models built for business! Granite 3.0 trained on 12 trillion+ tokens on data from 12 natural languages and 116 programming languages. ⚡ ⚡ Granite will become the default model of IBM Consulting Advantage, an AI-powered delivery platform used by IBM's consultants. #IBM #IBMConsulting #AI #IBMGranite #genAI #ConsultingAdvantage

This is such a great announcement to see. Excellent benchmark results with smaller models showing excellent accuracy and performance at a fraction of the cost (and energy footprint) compared to some of the larger models out there. New Guardian model which helps improve on AI Safety. And of course the continued focus on transparency. It is always better to know what data the model has been trained on and how!

IBM Granite 3.0 models deliver state-of-the-art performance relative to model size while maximizing safety, speed and cost-efficiency for enterprise use cases. Looking forward to leveraging these models to accelerate code generation in Watson Code Assistant ! ! !

The era off small, high performing models is here! Combined with enterprise data, they offer task-specific performance that rival large models at a fraction of the cost.    With Granite 3.0, Indian customers can scale trusted #GenAI using open and performant models that are tailored for specific business needs.    #IBM #AI

💡 Introducing IBM Granite 3.0 💡 Today, IBM announced the debut of our most advanced family of AI models yet, Granite 3.0. These highly performant LLMs are cost-efficient and transparent, released under a fully permissive Apache 2.0 license. This third-generation of Granite models will empower developers to build AI applications with a new level of trust and scalability, for a wide variety of enterprise use cases. At IBM, we believe that the future of AI is open, and these models are a fantastic representation of our ethos as we help to shape the most #innovative, #collaborative, #effective and #safe #future for AI — for #everyone. Granite 3.0 is just the beginning, and IBM is looking forward to advancing a stronger AI ecosystem that allows us to problem solve more efficiently than ever before.https://ibm.biz/BdavpB #ResponsibleAI #IBM #AI #IBMGranite #LLM #Apache #HuggingFace #Watson.ai

🚀 Introducing IBM's Granite 3.0 AI Models: Small, High-Performing, and Built for Business! 🚀 At #IBMTechXchange, we unveiled the latest Granite 3.0 models—compact, enterprise-grade AI models that are setting new benchmarks for performance, transparency, and safety. These models are purpose-built for business, offering high efficiency at a fraction of the cost of larger models. 🔹 What's New with Granite 3.0? Smaller models (8B, 2B) that deliver best-in-class performance for AI tasks like RAG, classification, and entity extraction. Fully permissive Apache 2.0 license—available on Hugging Face. New Granite Guardian models providing comprehensive safety guardrails for risk detection. 🔹 Key Features: On benchmarks, Granite 3.0 outperforms leading models from Meta and Mistral. Fine-tuned with enterprise data, enabling task-specific performance that matches or exceeds larger models—at a fraction of the cost. Optimized for low-latency applications and CPU-based deployments. 🔹 Scaling AI for Business: IBM’s Granite 3.0 models deliver trusted, cost-effective AI for enterprises, empowering businesses to innovate without complexity. Join us as we redefine AI for business! 💡 Learn more at our watsonx Rally on October 23 and explore the full potential of AI for business. #AI #LLM #Granite3 #EnterpriseAI #watsonx #IBMTechXchange

IBM may not be the first company that comes to mind when you think about AI, but we are changing the way business leverage AI. Last year we introduced #watsonx. This year in conjunction with Red Hat we introduced #InstructLab and now a new generation of Granite Models that outperforms the leading models in the market. We recognize that its not just about new AI technology, but unlocking business data to drive differentiation. Most enterprise data is on-prem and not public. With IBM Fusion HCI, we make running enterprise AI easy, reducing the time to get into production down to a matter of couple of days. At IBM, we have transformed into a Hybrid Cloud and AI company. How can we help you do the same. #letscreate