Multiverse Computing: CompactifAI nearly doubles Llama 3.3 performance on Intel Xeon 6
Multiverse Computing
Meta
Intel Corporation
Multiverse Computing's CompactifAI technology nearly doubles the performance of Meta's Llama 3.3 model on Intel Xeon 6 processors, achieving significant speedups without sacrificing accuracy. The solution uses quantum-inspired techniques to optimize AI inference.
Multiverse Computing has announced that its CompactifAI technology nearly doubles the performance of Meta's Llama 3.3 large language model on Intel Xeon 6 processors. The optimization achieves significant speedups in inference without sacrificing accuracy. CompactifAI uses quantum-inspired algorithms to compress and accelerate AI models. This development enables more efficient deployment of LLMs on standard server hardware.
- Сокращения
- LLM = Large Language Model — Большая языковая модель
Source: Meta AI (GNews) —
original