阿里云近日宣布，其通义千问系列再添新成员——320亿参数模型Qwen1.5-32B，该模型在性能、效率和内存占用之间取得了卓越的平衡，为业界和开发者提供了更具成本效益的解决方案。这一开源举措进一步丰富了通义千问的模型矩阵，目前该系列已公开了总计7款不同规模的大语言模型。
据透露，通义千问的开源模型在海内外开发者社区中受到了热烈反响，累计下载量已突破300万次，显示出其广泛的应用价值和影响力。这些模型的开源，不仅推动了人工智能技术的普惠，也为学术研究和商业创新提供了强大的工具支持。
通义千问先前已开源了涵盖5亿、18亿、40亿、70亿、140亿和720亿参数的6款模型，并且所有这些模型都已升级至最新的1.5版本，以确保用户能够享受到最先进的自然语言处理能力。这一系列的开源行动，无疑彰显了阿里云在人工智能领域的技术实力和开放合作的承诺，为全球AI生态的繁荣做出了积极贡献。
英语如下：
**News Title:** “Alibaba Cloud’s Qwen1.5-32B, a 320-billion-parameter model, ignites open-source fervor; leading innovation in the field”
**Keywords:** Qwen, Open-source model, Parameter upgrade
**News Content:** Alibaba Cloud recently announced the addition of a new member to its Qwen series – the Qwen1.5-32B model with 320 billion parameters. This model achieves an exceptional balance between performance, efficiency, and memory usage, offering a more cost-effective solution for the industry and developers. This open-source move further expands the Qwen model matrix, which now encompasses a total of seven large language models of varying scales.
It has been disclosed that Qwen’s open-source models have garnered significant enthusiasm in domestic and international developer communities, with cumulative downloads surpassing 3 million, demonstrating their wide-ranging application value and influence. By making these models open-source, they have not only promoted the democratization of AI technology but also equipped academic research and commercial innovation with robust tooling support.
Previously, Qwen open-sourced six models with parameter counts ranging from 500 million to 720 billion, all of which have been upgraded to the latest 1.5 version, ensuring users access to the most advanced natural language processing capabilities. These open-source initiatives underscore Alibaba Cloud’s technological prowess in AI and its commitment to collaboration, positively contributing to the global AI ecosystem’s prosperity.
【来源】https://www.leiphone.com/category/industrynews/ZF6wxlum2yVXJXzo.html
Views: 1