As society begins to grasp the immense potential of AI, leading applied research lab Prem AI is challenging the prevailing “bigger is better” narrative and leading the revolution in Small Language Models (SLMs). As tech giants continue to unveil massive language models with billions of parameters, Prem AI is redefining the future of AI by focusing on efficient, targeted, and highly customizable SLMs that are changing the way businesses and developers approach artificial intelligence.
The Power of Small Language Models
At the heart of Prem AI’s innovation is the recognition that, while Large Language Models (LLMs) have their place, they are often overkill for many real-world applications.
“We believe that the future of AI lies not in creating ever-larger models but in developing targeted, efficient solutions that can be widely deployed and easily customized,” says Simone Giacomelli, founder and CEO of Prem AI. “Our focus on SLMs isn’t about compromising capability; it’s about maximizing impact while minimizing resource requirements.”
The company’s flagship model, Prem-1B, exemplifies the potential of SLMs to deliver impressive performance in a compact package. With just 1 billion parameters, Prem-1B offers capabilities that were once thought exclusive to much larger models, all while maintaining efficiency and accessibility.
Prem AI’s unique approach has garnered significant attention and backing, having recently raised $14 million from prominent investors, including David Maisel (Marvel Studios), Fan Zhang (Sequoia) and Alan Lipschitz (Incubeta).
Customization: The Key to AI Success
Prem AI’s approach goes beyond just developing smaller models. In addition to developing an end-to-end AI engineering platform, the company has positioned itself as a leader in customizing and fine-tuning SLMs for specific industry needs. This approach allows businesses to harness the power of AI without the overhead of massive, general-purpose models. Through its research, Prem AI has found that, in some cases, smaller models can outperform larger, general-purpose models in accuracy and speed, all while maintaining strict data privacy standards.
“Our ability to rapidly iterate and fine-tune models for specific use cases sets us apart,” explains Filippo Peddrazini, Prem AI’s CTO. “We’re not just providing a product; we’re offering a partnership that evolves with our clients’ needs.”
Innovations in SLM Technology
Prem AI’s leadership in the SLM space is driven by continuous innovation. The company has made significant progress in several key areas:
- Advanced Retrieval-Augmented Generation (RAG): Prem AI is revolutionizing enterprise data management with its new Retrieval-Augmented Generation (RAG) techniques integrated into its platform. This innovative technology enables enterprises to tailor their private SLMs to seamlessly access and integrate rich external knowledge bases, as well as internal PDFs and .doc files. By using RAG, businesses can improve their data utilization and ensure that critical information is easily accessible and effectively utilized within their operations.
- Efficient Tool Calling: Tool calling enables SLMs to interact with product databases and CRM systems to retrieve detailed information about specific products. This results in personalized recommendations tailored to individual customer preferences. By harnessing this innovative approach, e-commerce businesses can improve the user experience and drive greater customer satisfaction through more relevant and timely product suggestions.
- Multi-Modal Integration: Prem AI is at the forefront of integrating SLMs with other AI modalities, such as computer vision and speech recognition, creating compact yet powerful multi-modal AI systems. Multimodal SLMs can generate a wide variety of content, including text, images, audio, and video. This content can be easily integrated into your applications. These models use cost-effective, nuanced, and context-aware AI to process and understand different types of sensory input.
Democratizing AI Through Open Source
In line with its mission to democratize AI, Prem AI has significantly contributed to the open-source community. The release of the Prem-1B base model and its fine-tuned chat variant has empowered developers and researchers worldwide to build upon and improve SLM technology.
“We believe that open collaboration is key to advancing the field of AI,” says Nicola Sosio, Lead of Open Source Initiatives at Prem AI. “By sharing our models and research, we’ve fostered an ecosystem where innovation thrives. With thousands of developers actively using Prem and dozens of businesses joining the fold, we’ve also received enthusiastic feedback from AI conferences around the world and see our impact continue to grow in the developer community.”
Looking to the Future
As Prem AI continues to push the boundaries of what’s possible with SLMs, the company is also
exploring new frontiers in AI development. Its current research focuses on further reducing model size while maintaining performance, developing more energy-efficient training methods, and creating models that can learn and adapt in real time.
“We’re just scratching the surface of what’s possible with SLMs,” Simone says. “Our vision is a future where AI is not just a powerful tool for big tech, but a vision of the future where millions, even billions, of personalized models are tailored for every institution, enterprise, and individual, enhancing every aspect of our lives and businesses.”
As the AI landscape continues to evolve, Prem AI stands at the forefront, not just adapting to change but actively shaping the future of the field. Through its innovative approach to SLMs, commitment to open source, and focus on real-world applications, Prem AI is proving that when it comes to AI, sometimes less really is more.
