谷歌宣布推出全球性能最强大、轻量级的开源模型系列Gemma，该系列分为2B（20亿参数）和7B（70亿）两种尺寸版本。2B版本甚至可在笔记本电脑上运行，这标志着人工智能模型在轻量化和性能方面取得了重大突破。
Gemma模型采用了与Google DeepMind开发的Gemini模型相同的研究和技术，专为负责任的AI开发而打造。在18个语言理解、推理、数学等关键基准测试中，Gemma在11个测试分数上超越了Meta Llama-2等更大参数的开源模型，展现了其卓越的性能。
谷歌的这一举措不仅推动了开源模型的技术发展，也为全球开发者提供了强大的工具，有助于促进人工智能技术的广泛应用和创新。
英文标题：Google Unveils Open-Source Gemma Models, Breaking Performance Barriers
英文关键词：Open-Source Models, Performance Leap, Responsible AI Development
英文新闻 content：
Google has announced the launch of the open-source model series Gemma, which is the most powerful and lightweight in the world. The series includes two versions: 2B (20 billion parameters) and 7B (70 billion), with the 2B version capable of running on a laptop, marking a significant breakthrough in both lightweight and performance aspects of AI models.
The Gemma models utilize the same research and technology as the Gemini models developed by Google DeepMind, specifically designed for responsible AI development. In 18 key benchmarks for language understanding, reasoning, and mathematics, Gemma outscored larger parameter open-source models like Meta Llama-2 in 11 tests, demonstrating its exceptional performance.
Google’s move not only propels the technological advancement of open-source models but also provides a powerful tool for developers worldwide, which is expected to foster widespread application and innovation in AI technology.
【来源】https://www.tmtpost.com/6944205.html
Views: 1