** MiniMax 稀宇科技发布万亿参数 MoE 模型：abab 6.5系列崭新登场 **
今日，国内知名科技公司 MiniMax 稀宇科技宣布推出其最新一代大规模预训练模型——abab 6.5 系列。这一系列包括两个模型：abab 6.5 和 abab 6.5s，标志着公司在人工智能领域的又一重大突破。据官方介绍，abab 6.5 模型拥有万亿级别的参数量，能够处理长达 200,000 个 tokens 的上下文信息，展现了强大的语言处理能力。
与此同时，abab 6.5s 作为该系列的轻量化版本，虽然同样采用了与 abab 6.5 相同的训练技术和数据集，但其效率更高，能够在短短一秒钟内处理近 30,000 字的文本，展现了极高的实时处理速度。这一性能的提升，使得 abab 6.5s 在应对大规模文本处理任务时，具有显著的竞争优势。
在各类核心能力的测试中，abab 6.5 系列表现出色，其性能已开始逼近全球顶尖的大型语言模型，如 OpenAI 的 GPT-4、Google 的 Claude-3 以及阿里云的 Gemini-1.5。这一成就不仅彰显了 MiniMax 稀宇科技在人工智能研发领域的深厚实力，也预示着中国公司在全球语言模型竞赛中正逐步崭露头角。
MiniMax 稀宇科技的这一创新发布，无疑为全球人工智能研究和应用领域注入了新的活力，期待abab 6.5 系列模型能在实际应用中带来更为智能、高效的语言理解和生成体验。
英语如下：
**MiniMax Unveils Trillion-Parameter MoE Giant: abab 6.5, Pushing the Limits of Global Language Models**
Keywords: MiniMax, trillion-parameter model, abab 6.5
**MiniMax XinYu Tech Launches Trillion-Parameter MoE Model: The abab 6.5 Series**
Renowned domestic tech company MiniMax XinYu Tech has announced the release of its latest generation of large-scale pre-training models — the abab 6.5 series. The series consists of two models, abab 6.5 and abab 6.5s, marking a significant milestone in the company’s artificial intelligence endeavors. According to official statements, the abab 6.5 model boasts a trillion-parameter scale and can handle contextual information of up to 200,000 tokens, demonstrating formidable language processing capabilities.
Simultaneously, the lightweight abab 6.5s, leveraging the same training techniques and dataset as abab 6.5, operates with increased efficiency, processing nearly 30,000 characters of text within a second, showcasing exceptional real-time processing speed. This performance enhancement gives abab 6.5s a competitive edge in handling large-scale text processing tasks.
In benchmark tests, the abab 6.5 series has demonstrated outstanding performance, approaching the capabilities of top global large language models such as OpenAI’s GPT-4, Google’s Claude-3, and Alibaba Cloud’s Gemini-1.5. This achievement underscores MiniMax XinYu Tech’s profound expertise in AI research and signals the rising prominence of Chinese companies in the global language model race.
MiniMax XinYu Tech’s innovative launch promises to infuse new vitality into the global AI research and application landscape. The abab 6.5 series is anticipated to deliver smarter and more efficient language understanding and generation experiences in practical applications.
【来源】https://mp.weixin.qq.com/s/xBoAP-6fZVQA9cEWT8gyfw
Views: 2