Multiverse Computing: CompactifAI nearly doubles Llama 3.3 performance on Intel Xeon 6

Multiverse Computing announced that its neural network compression technology CompactifAI can nearly double the performance of the large language model Llama 3.3 when running on Intel Xeon 6 server processors. The company claims that CompactifAI uses quantum-inspired and tensor compression methods to reduce model size and accelerate inference while maintaining high accuracy. Testing showed that on the Intel Xeon 6 platform with 288 cores, task execution time is reduced by almost 50%, making the model more efficient for deployment in data centers. Intel and Multiverse Computing emphasized that the solution requires no specialized hardware and runs on standard CPUs.