谷歌近日宣布了一项重大技术创新，推出了开源大模型系列Gemma，这是一款在性能和轻量化上均达到新高度的AI模型。Gemma分为2B和7B两个版本，其中2B参数量的模型竟然可以在普通的笔记本电脑上流畅运行，这无疑为AI研究和开发带来了极大的便利。
据谷歌透露，Gemma受益于与Gemini模型相同的研究理念和技术，由Google DeepMind和谷歌的多支团队联合打造，旨在推动负责任的AI发展。在确保效率的同时，Gemma在语言理解、推理和数学等多个关键基准测试中表现出色，尤其在18项评估中，有11项得分超过了Meta的Llama-2等更大规模的开源模型，展现了其强大的智能潜力。
这一创新标志着谷歌在AI领域的又一重要突破，Gemma的开源特性将为全球的开发者和研究者提供一个更强大、更易访问的平台，进一步推动AI技术的民主化和进步。谷歌的这一举措无疑将激励更多的创新者参与到AI的研发中，共同构建更加智能和负责任的未来。
英语如下：
News Title: “Google Unveils Open-Source Gemma, a Lightweight AI Powerhouse Breaking Performance Records!”
Keywords: Google release, Gemma model, performance surpassing
News Content:
Title: Google’s Open-Source Gemma Shines, Outperforming Meta’s Llama-2 in Lightweight Brilliance
Google recently announced a major technological innovation with the release of the open-source Gemma series, an AI model that sets new standards in performance and compactness. Available in 2B and 7B parameter versions, the 2B model impressively runs smoothly on a standard laptop, revolutionizing convenience for AI research and development.
Derived from the same research principles and techniques as the Gemini model, Gemma is a collaborative effort between Google DeepMind and multiple Google teams, aiming to advance responsible AI. While maintaining efficiency, Gemma excels in language understanding, reasoning, and math across several key benchmarks. Remarkably, it outperforms larger open-source models like Meta’s Llama-2 in 11 of 18 assessments, demonstrating its immense intellectual potential.
This innovation signifies another significant milestone for Google in the AI realm. By making Gemma open-source, Google is providing a more powerful and accessible platform for developers and researchers worldwide, further democratizing AI advancements and encouraging more innovators to contribute to a smarter, more responsible future.
【来源】https://www.tmtpost.com/6944205.html
Views: 7