Latest 6G and AI news, analysis, and industry updates.
Liquid AI's 350-million-parameter LFM2.5-350M demonstrates reliable data extraction and tool calling, suggesting compact models can run agentic workflows in memory- and latency-constrained environments.
A wave of sub-10B parameter models from Mistral, Microsoft, and academic labs are outperforming 100B+ models on domain-specific benchmarks. This shift is redefining the economics of AI deployment and challenging the 'bigger is better' paradigm.
On-device AI inference is exploding, powered by Apple's Neural Engine, Qualcomm's AI Engine, and specialized NPUs. From smartphones to industrial sensors, edge AI is enabling real-time intelligence without cloud dependency, transforming privacy, latency, and cost economics.