**MiniMax 发布万亿参数 MoE 模型,abab 6.5 系列引领语言处理新纪元**
今日,全球知名科技企业 MiniMax 稀宇科技宣布了一项重大突破,正式推出了其新一代大规模预训练模型——abab 6.5 系列。该系列包括两个模型,即 abab 6.5 和 abab 6.5s,均在语言处理能力上达到了前所未有的水平。
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,展现了强大的语境理解与处理能力。而其轻量级版本 abab 6.5s,虽然同样基于相同的训练技术和数据,但效率更胜一筹,同样支持 200k tokens 的上下文长度,能在短短1秒内处理近 30,000 个汉字的文本,速度之快令人惊叹。
在多项核心能力测试中,abab 6.5 系列模型的表现已接近目前全球公认的领先大语言模型,如 GPT-4、Claude-3 和 Gemini-1.5 等。这一成就标志着 MiniMax 在人工智能语言模型领域的领先地位,同时也预示着未来自然语言处理技术将有更广阔的应用前景。
MiniMax 稀宇科技的这一创新举措,无疑将对全球的科研、教育、媒体及各行各业产生深远影响,为大数据时代的信息处理与理解提供更为高效和精准的工具。随着 abab 6.5 系列模型的发布,我们期待看到更多基于此技术的创新应用,推动人工智能与人类社会的深度融合。
英语如下:
**News Title:** “MiniMax Sets a Record with Trillion-Parameter MoE Model abab 6.5, Challenging the Apex of Global Language Models”
**Keywords:** MiniMax, trillion parameters, MoE model
**News Content:**
Today, renowned global tech firm MiniMax Xiyu Tech announced a groundbreaking advancement with the release of its next-generation large-scale pre-training model, the abab 6.5 series. Consisting of two models, abab 6.5 and abab 6.5s, both have achieved unparalleled capabilities in language processing.
According to official statements, the abab 6.5 model boasts a trillion-parameter count, enabling it to handle contextual information of up to 200,000 tokens, demonstrating exceptional contextual understanding and processing power. Its lightweight variant, abab 6.5s, leveraging the same training techniques and data, excels in efficiency. It can process nearly 30,000 Chinese characters within a single second, handling 200k tokens’ worth of context, showcasing remarkable speed.
In several core capability tests, the abab 6.5 series models have performed on par with the world’s leading large language models, such as GPT-4, Claude-3, and Gemini-1.5. This milestone underscores MiniMax’s dominance in the AI language model arena and foreshadows an even broader application horizon for natural language processing technology.
MiniMax Xiyu Tech’s innovative move is set to have profound implications across global research, education, media, and various industries, offering more efficient and precise tools for information processing and understanding in the era of big data. With the launch of the abab 6.5 series, we anticipate witnessing more innovative applications stemming from this technology, further integrating artificial intelligence into human society.
【来源】https://mp.weixin.qq.com/s/xBoAP-6fZVQA9cEWT8gyfw
Views: 1