The article reports on a significant advancement in AI model efficiency, detailing a new technique that drastically reduces the computational resources required for training and inference. Researchers have developed a method that compresses large language models without sacrificing performance, potentially making powerful AI more accessible. The approach involves a novel pruning algorithm that identifies and removes redundant parameters within neural networks. Early benchmarks show models reduced by up to 60% in size while maintaining over 95% of their original accuracy on standard tasks. This development could lower the barrier to entry for deploying advanced AI in resource-constrained environments, from mobile devices to smaller research labs. The full implications for the industry and further technical details are available in the complete article at https://technologyreview.com/2024/05/ai-model-compression-breakthrough.