Small Models & Edge Efficiency
Articles on parameter-efficient AI models, small language models (SLMs), and task-specific architectures optimized for edge deployment on smartphones, IoT devices, robotics, and local hardware — covering efficiency breakthroughs that enable capable AI without cloud dependency.
PrismML's Tiny Models Force a Rethink of AI Edge Infrastructure
PrismML's ternary quantization compresses a 27-billion-parameter reasoning model under six gigabytes, promising agentic AI on laptops and phones — and forcing a rethink of how AI edge infrastructure gets built, funded, and used.
Optimizing Models for AI Edge Infrastructure: Practical Quantization Strategies
A practical guide to model quantization for edge deployment, covering reduced model size, faster inference via integer arithmetic, and bitsandbytes integration.
Why Liquid AI's 230M-Parameter Model Beats Bigger Systems at the Jobs That Matter
Liquid AI's LFM2.5-230M demonstrates that the future of production AI isn't about scale — it's about task-specific optimization. At just 230 million parameters, this model beats models 4× its size at structured data extraction and runs on smartphones, laptops, and robotics hardware without cloud dependency.