How Microsoft & IBM’s SLMs Are Making AI More Sustainable
Is bigger always better? In the relentless pursuit of more powerful artificial intelligence, tech giants Microsoft and IBM are challenging this long-held notion. The companies are leading a shift toward smaller, more focused AI models to provide greater energy efficiency. These small language models (SLMs) demonstrate that powerful performance doesn’t demand excessive computational resources and energy expenditure.
Making Sustainable AI
The traditional AI development race has been characterized by relentlessly pursuing ever-larger models, consuming massive computational resources and energy. However, both IBM and Microsoft are now demonstrating that smaller, more focused AI models can deliver powerful performance while significantly reducing environmental and economic costs.
IBM’s Granite: Compact and Capable
IBM’s latest Granite 3.2 models further highlight the trend. This new approach focuses on compact systems designed for specific business applications that:
- Reduce computational requirements by up to 30% in the Guardian safety models
- Process complex document understanding tasks with minimal resource consumption
- Offer optional “chain of thought” reasoning to optimize computational efficiency
The TinyTimeMixers models are capable of two-year forecasting with fewer than 10 million parameters – a stark contrast to the hundreds of billions of parameters in traditional large language models.
Microsoft’s Phi-4: Multimodal Efficiency
Microsoft’s Phi-4 family is taking a similar approach. Microsoft has introduced two groundbreaking models:
- Phi-4-multimodal: A 5.6B parameter model that simultaneously processes speech, vision, and text
- Phi-4-mini: A 3.8B parameter model optimized for text-based tasks
These models are designed for compute-constrained environments, making them ideal for integration into smartphones, vehicles, and other devices with limited computational resources.
“Phi-4-multimodal marks a new milestone in Microsoft’s AI development as our first multimodal language model,” says Weizhu Chen. “By leveraging advanced cross-modal learning techniques, this model enables more natural and context-aware interactions, allowing devices to understand and reason across multiple input modalities simultaneously. Whether interpreting spoken language, analyzing images, or processing textual information, it delivers highly efficient, low-latency inference – all while optimizing for on-device execution and reduced computational overhead.”
Beyond performance: A sustainable vision
Both companies emphasize that the future of AI isn’t about raw computational power but about efficiency, integration, and real-world impact.
“The next era of AI is about efficiency, integration, and real-world impact – where enterprises can achieve powerful outcomes without excessive spend on compute,” says Sriram Raghavan, Vice President of IBM AI Research.
The key sustainability benefits include:
- Reduced energy consumption: Smaller models require significantly less energy to train and operate.
- Lower carbon footprint: Decreased computational needs translate to reduced greenhouse gas emissions.
- Increased accessibility: More affordable AI solutions for smaller organizations
- Flexible deployment: Ability to run advanced AI on edge devices and in resource-constrained environments.
