Analyst(s): Nick Patience
Publication Date: October 22, 2024
IBM has unveiled its Granite 3.0 family of AI foundation models, prioritizing efficiency and performance over sheer size. The models are designed for various applications, including safety-focused tasks and low-latency requirements on smaller devices. IBM is also integrating the models into its consulting services to deliver business value.
What is Covered in this Article:
- IBM’s new Granite 3.0 models prioritize efficiency and performance, aiming to deliver value at a lower cost than larger models.
- IBM emphasizes its support for open source by licensing the models under the Apache 2 license.
- New Granite models include versions specifically designed for safety and low-latency applications. Granite Guardian models focus on implementing safety guardrails, while Granite Mixture-of-Experts models target low-latency tasks on smaller devices.
- IBM sees Granite models as a key component of its broader AI strategy, integrating them into its consulting services and internal operations to drive practical business value for clients.
The News: IBM has unveiled its Granite 3.0 family of AI foundation models designed to bring AI capabilities to businesses while prioritizing efficiency and performance.
IBM Places Bet on Efficiency with New Granite 3.0 Foundation Model Family
Analyst Take: IBM’s announcement of Granite 3.0, its new foundational model, represents a significant step in the company’s AI strategy. The models include general purpose language models, models focused on guardrails and safety, and models aimed at low latency applications running on small devices, aimed in part at AI inference.
Efficiency and Performance as Key Drivers
IBM doesn’t believe that winning the race to build ever-larger models is where it should be placing its bets. Instead, the Granite 3.0 models are specifically designed with a focus on efficiency and performance. But it does believe that large language models represent a new programming model so it’s important for IBM to participate in that fundamental shift. This focus on efficiency is further reinforced by the company’s exploration of smaller models, highlighting a shift toward optimizing AI for specific tasks and reducing reliance on resource-intensive large language models. IBM released various benchmarks showing Granite model comparison favorably against similar size and even larger models from companies such as Meta and Mistral. IBM gave an example at its analyst forum in New York last week of a Granite model in use at a large financial services company, claiming not only does the model outperform OpenAI’s GPT-4 Turbo but it did so at a considerably lower cost, in terms of costs per million tokens.
Guardrails and Inference
The new Granite Guardian 3.0 8B and Granite Guardian 3.0 2B models are aimed specifically at organizations that want to implement safety guardrails. They provide risk and harm detection capabilities to identify potential risks in both user prompts and AI-generated responses.
The Granite Mixture-of-Experts models are aimed at low latency applications on small devices run on CPUs. These devices are where a lot of AI inference will be happening. IBM claims they can achieve performance levels comparable to much larger models while consuming significantly less computational resources.
Open Source and Governance
The entire family of Granite 3.0 models are available under the Apache 2 permissive use license. IBM has had a deep commitment to open source for more than 20 years, starting with Linux and then buying Red Hat. IBM’s watsonx.governance tool designed to address concerns related to bias, explainability, and risk in AI models is a real differentiator for the company based on decades of governance work.
Addressing Real-World Challenges
IBM’s AI strategy extends beyond just developing advanced models. The company is actively addressing real-world challenges by integrating AI into its consulting services and internal operations. For instance, IBM Consulting has incorporated AI into its platform, Consulting Advantage, to enhance consultant efficiency and deliver better client outcomes. This practical approach to AI implementation demonstrates IBM’s commitment to driving tangible business value for its clients.
What to Watch:
- Market Reception: It will be crucial to observe how the market receives Granite 3.0 and its performance claims. After all, there are plenty of models and new announcements are expected from IBM’s main enterprise AI rivals on a regular basis.
- Competition: The AI landscape is evolving rapidly, with major players such as Microsoft, Google, and AWS investing heavily in AI. It remains to be seen how IBM will differentiate itself in this competitive market as it is not a hyperscale cloud provider.
- IBM Consulting provides IBM with a different type of story to tell compared to the cloud companies and its use of the Granite models should drive efficiencies for IBM Consulting and its customers.
See the complete press release on Granite 3.0 on the IBM company website.
Disclosure: The Futurum Group is a research and advisory firm that engages or has engaged in research, analysis, and advisory services with many technology companies, including those mentioned in this article. The author does not hold any equity positions with any company mentioned in this article.
Analysis and opinions expressed herein are specific to the analyst individually and data and other information that might have been provided for validation, not those of The Futurum Group as a whole.
Other Insights from The Futurum Group:
Google Cloud’s AI-First Vision: Empowering Businesses for the Generative AI Era
IBM Reports Q2 2024 Financial Results: Key Insights and Performance Review
IBM’s New Telum II Processor – A Strategic Advancement in Transactional AI
Author Information
Nick Patience is VP and Practice Lead for AI Platforms at The Futurum Group. Nick is a thought leader on AI development, deployment, and adoption - an area he has researched for 25 years. Before Futurum, Nick was a Managing Analyst with S&P Global Market Intelligence, responsible for 451 Research’s coverage of Data, AI, Analytics, Information Security, and Risk. Nick became part of S&P Global through its 2019 acquisition of 451 Research, a pioneering analyst firm that Nick co-founded in 1999. He is a sought-after speaker and advisor, known for his expertise in the drivers of AI adoption, industry use cases, and the infrastructure behind its development and deployment. Nick also spent three years as a product marketing lead at Recommind (now part of OpenText), a machine learning-driven eDiscovery software company. Nick is based in London.
