How SLMs made AI more compact and powerful in 2024

If 2023 was all about Large Language Models (LLMs), then in 2024 Small Language Models (SLMs) have significantly enhanced AI’s efficiency and accessibility across various sectors. Their streamlined architectures enable deployment in resource-constrained environments, making them ideal for mobile devices and Internet of Things (IoT) applications. This shift has led to the development of tools like Microsoft Translator Lite, which offers real-time offline translations, and SLM-powered smart assistants that process voice commands directly on wearables, enhancing user experience while conserving energy.

Beyond mobile and IoT, SLMs have driven advancements in energy-efficient AI applications. Models such as GreenBERT have reduced data center power consumption by 30%, while Tesla’s PowerWall AI optimizes energy storage in smart homes, contributing to a 20% increase in energy savings. In agriculture, SLM-powered sensors provide real-time crop monitoring, improving yields by an average of 15%. 

Also read: SLM vs LLM: Why smaller Gen AI models are better

These developments underscore SLMs’ pivotal role in creating sustainable, efficient, and accessible AI solutions in 2024. Let’s recap some of the top SLM highlights.

1) Mobile and IoT applications driven by SLMs

Small Language Models (SLMs) have become integral to mobile and IoT ecosystems. Their lightweight architecture and localised capabilities cater to applications that demand low latency and minimal energy consumption, making them ideal for offline and on-the-go use cases. Here are some of the standout applications:

Microsoft Translator Lite: Real-time multilingual translations offline

Microsoft Translator Lite leverages compact SLMs to provide real-time translations without requiring an internet connection. In 2024, the application reached over 70 million monthly active users globally, driven by its inclusion of support for emerging languages like Swahili and Tamil. Microsoft reported that Translator Lite has improved global productivity by reducing communication barriers for over 40% of small-to-medium enterprises using the tool in regions with limited internet access. 

Also read: Microsoft Phi-4: How to use, AI applications, everything you should know

With support for over 50 languages, including emerging regional dialects, the tool caters to frequent travellers, remote workers, and multilingual professionals. According to Microsoft’s internal research, the app has reduced data usage by 80%, which is crucial in regions where mobile data is expensive or unreliable. This innovation addresses the persistent challenge of ensuring accessibility across language barriers while minimising data usage, contributing to a 25% increase in adoption compared to cloud-dependent alternatives.

SLM Smart Assistants: Voice commands in wearables without connectivity

SLM-powered smart assistants have redefined wearable technology by enabling offline voice command processing. Fitness trackers and smartwatches can now execute tasks like setting reminders, tracking activities, and controlling music playback without relying on cloud-based services. These localised capabilities ensure faster response times and greater privacy, as sensitive voice data remains on the device. The integration of SLMs has also reduced the computational burden by 30%, extending battery life for wearables by an average of 20 hours per charge. This advancement is particularly significant as wearable shipments surpassed 250 million units globally in 2024.

PocketLens AI: Summarising printed books via smartphone camera

PocketLens AI utilises lightweight models to analyse and summarise content captured through a smartphone’s camera. Aimed at students and professionals, this application simplifies research by extracting key points from textbooks and documents. With support for over 30 languages and specialised features for STEM texts, the app has been downloaded over 10 million times since its launch in early 2024. Its ability to integrate seamlessly with note-taking apps and digital libraries makes it a valuable tool for enhancing productivity.

Instant Transcriptions: Fast speech-to-text conversions with LiveScribe

SLM-driven tools like LiveScribe have made speech-to-text transcription faster and more accessible. These tools convert spoken words into text in real time, catering to journalists, content creators, and corporate professionals. LiveScribe’s 2024 update introduced context-aware tagging and multilingual transcription support for 15 additional languages, increasing its user base by 40%. With a transcription accuracy rate of 96% in offline mode, LiveScribe has become an indispensable tool for users working in connectivity-challenged environments.

IoT Ecosystems: Local voice commands for appliance control

SLMs play a critical role in expanding the functionality of IoT ecosystems. They enable users to control multiple smart appliances locally via voice commands, eliminating the need for cloud-based processing. For instance, users can manage lighting, thermostats, and security cameras using a centralised hub powered by an SLM.

Also read: AI agents explained: Why OpenAI, Google and Microsoft are building smarter AI agents

This architecture reduces latency by 25%, enhances security by keeping data localised, and ensures uninterrupted service even during network outages. In 2024, the global smart home market grew by 18%, with SLM-driven IoT solutions accounting for a significant share of new deployments.

2) Energy Efficiency powered by SLMs

SLMs are also driving significant advancements in energy-efficient AI applications. Their streamlined architectures reduce the computational load, making them suitable for devices and systems prioritising sustainability. Here are some of the key developments in 2024:

GreenBERT: Reduced energy consumption in enterprise data centres

GreenBERT is an SLM variant designed specifically for data centers, optimising natural language processing tasks with minimal energy use. In 2024, it achieved a 30% reduction in power consumption compared to traditional models, saving enterprises an estimated $1 billion annually in energy costs. By employing quantisation techniques and hardware-efficient designs, GreenBERT has become a cornerstone for businesses aiming to lower operational expenses while meeting stringent environmental regulations.

Tesla PowerWall AI: Energy storage and distribution optimisation

Tesla’s PowerWall AI incorporates SLMs to optimise energy storage and distribution in smart homes. By analysing energy consumption patterns, solar panel output, and local electricity tariffs, the system recommends cost-effective usage strategies. Tesla reported a 20% increase in energy savings for PowerWall users in 2024, reflecting the system’s ability to adapt to real-time changes in energy supply and demand. This innovation aligns with the broader push for sustainable living, with over 500,000 units sold this year alone.

Low-Energy AI Sensors: Crop monitoring in remote fields

SLM-powered sensors are revolutionising agriculture by enabling real-time crop monitoring in remote locations. These sensors analyse soil moisture, temperature, and weather conditions to provide actionable insights to farmers. In 2024, such sensors covered over 2 million hectares globally, helping farmers improve crop yields by an average of 15%. Their ultra-low power consumption ensures extended operational periods using solar or battery power, making them indispensable for rural and underdeveloped regions.

EcoPhone AI: Ultra-low-power AI chips in smartphones

EcoPhone AI represents a new category of smartphones equipped with ultra-low-power AI chips driven by SLMs. These devices handle tasks like voice recognition, predictive typing, and photo optimisation locally, significantly extending battery life. With energy consumption reduced by up to 40%, EcoPhone AI models have gained significant traction in 2024, especially among environmentally conscious consumers. Over 5 million units have been sold globally within six months of launch.

Efficient Electric Cars: Optimized Energy Output in EVs

SLMs are now embedded in electric vehicle (EV) systems to optimise energy output dynamically. In 2024, approximately 40% of newly launched EV models adopted SLM-powered energy management systems, enhancing their range and performance. User reviews consistently highlight the technology’s ability to adapt to diverse driving conditions, with many citing a noticeable improvement in energy efficiency. For instance, mid-tier EVs equipped with SLMs reported an average 15% increase in range, with some models achieving up to 20% under ideal conditions. 

Also read: KOGO OS aims to popularise LLM-agnostic approach to AI: Here’s how

This has played a pivotal role in boosting global EV sales by 25% this year. By analysing factors like driving habits, road conditions, and battery health, these models ensure efficient energy utilisation. In 2024, mid-tier EVs equipped with SLMs reported a 15% increase in range, addressing key consumer concerns about EV adoption. This advancement has contributed to a 25% growth in global EV sales, solidifying SLMs’ role in the automotive sector.

3) Consumer Accessibility of SLMs

SLMs have democratised access to AI-driven tools, making them affordable and intuitive for a wider audience. Their compact size and efficiency enable seamless integration into consumer devices and services. Below are some noteworthy advancements:

Budget AI Phones: Entry-level models with voice recognition

In 2024, brands like Nokia introduced entry-level smartphones equipped with SLM-based voice recognition. Priced under $100, these phones quickly gained traction in emerging markets, enabling hands-free navigation, call management, and text dictation for first-time smartphone users. By mid-2024, these devices had sold over 10 million units globally, reflecting their affordability and utility in narrowing the digital divide. 

Also read: LLM to RAG: Decoding AI jargon matters, here’s why

According to industry reports, entry-level AI phones accounted for 15% of total smartphone sales in developing regions, underscoring their transformative potential in bridging technology gaps. These devices offer hands-free navigation, call management, and text dictation, catering to users in emerging markets. Priced under $100, these phones have brought AI capabilities to over 10 million new users, narrowing the digital divide. This democratisation of technology is particularly impactful in regions like Africa and South Asia, where smartphone penetration remains a priority for development.

AI Text Simplifiers: Enhanced readability of web articles

SLMs have been instrumental in developing tools that simplify complex web content into more digestible formats. These applications analyse text structure and terminology to rewrite articles at varying reading levels, ensuring accessibility for diverse audiences, including non-native speakers and individuals with learning difficulties. Browser plugins featuring this technology have seen a 50% increase in downloads in 2024, underscoring the demand for inclusive online experiences.

Language-Specific Apps: Predictive typing for regional languages

SLMs power AI keyboards that support predictive typing in multiple regional languages. By understanding context and local linguistic nuances, these keyboards provide accurate word suggestions and grammar corrections. In India alone, where over 20 regional languages are commonly used, such apps have facilitated a 30% increase in typing efficiency, making digital communication more accessible to millions of users.

Voice Memo Organizers: Intelligent sorting and summarisation

Voice memo applications have been enhanced with SLMs to sort, tag, and summarise audio notes automatically. Users can search for keywords or themes within their recordings, making it easier to retrieve relevant information. Tools like MemoGen have grown in popularity, with downloads exceeding 1 million in 2024. This functionality caters to professionals and students who rely on voice memos for capturing spontaneous ideas or detailed discussions.

SLM Navigation: Offline map services with dynamic updates

Offline navigation apps now incorporate SLMs to provide dynamic route updates based on real-time factors like road closures or traffic conditions. These models ensure that users receive accurate and up-to-date directions even without an internet connection. In 2024, such solutions have become essential for travellers, with apps like NavigateSLM seeing a 35% growth in user base. This capability is particularly valuable in remote areas and for adventurers navigating challenging terrains.

Also read: How LLMs revolutionised our AI journey in 2024