【北京，2023年4月2日】国内科技巨头APUS与战略伙伴新旦智能携手，共同创造了历史性的突破，成功开源了国内首个基于千亿级MoE架构的大模型。这一名为APUS-xDAN大模型4.0的创新成果，不仅在GitHub上引起了广泛关注，更标志着中国在大模型研发领域迈出了重要一步。
APUS-xDAN大模型4.0采用了先进的Mixture of Experts（MoE）架构，其独特之处在于能够在4090等相对低端的算力芯片上高效运行，实现了技术普惠的目标。这一创新不仅降低了使用高参数模型的门槛，也为行业内的广大开发者提供了更广阔的应用空间。据官方数据显示，APUS-xDAN大模型4.0在综合性能上已达到全球知名的GPT-4的90%，展现出强大的计算效能和应用潜力。
这一开源项目的发布，无疑将推动中国乃至全球的AI技术研发和应用进入新的阶段。APUS与新旦智能的深度合作，不仅展现了中国科技企业的创新实力，也为全球AI社区贡献了宝贵的资源，进一步推动了大模型技术的普惠和进步。
APUS公众号发布的消息强调，此次开源行动旨在促进技术交流与合作，鼓励更多开发者参与到大模型的优化和创新中，共同推动人工智能技术的边界。这一开源大模型的诞生，无疑将为AI领域的研究和应用打开新的可能，期待未来能够见证更多基于此模型的创新应用和服务的诞生。
英语如下：
**News Title:** “APUS and Xin Dan Intelligence Create History: Launching China’s First Open-Source Trillion-Level MoE Large Model for Low-End Chips”
**Keywords:** APUS, Trillion MoE, Open-Source Large Model
**News Content:**
**Beijing, April 2, 2023** — Tech giant APUS, in collaboration with strategic partner Xin Dan Intelligence, has made a historic breakthrough by successfully launching China’s first open-source large model based on a trillion-scale Mixture of Experts (MoE) architecture. The groundbreaking innovation, called APUS-xDAN Large Model 4.0, has garnered widespread attention on GitHub and signifies a significant step forward for China in large model development.
APUS-xDAN Large Model 4.0 employs cutting-edge MoE architecture, notable for its efficient operation on relatively low-end computing chips like the 4090, thus promoting technological inclusiveness. This innovation lowers the barrier to using high-parameter models and expands the application landscape for developers across the industry. Official data indicates that APUS-xDAN 4.0 achieves 90% of the overall performance of the globally renowned GPT-4, demonstrating its robust computational efficiency and potential.
The release of this open-source project is set to propel AI research and application in China and globally into a new era. The deep collaboration between APUS and Xin Dan Intelligence showcases the innovative strength of Chinese tech companies and contributes valuable resources to the global AI community, further advancing the democratization and progress of large model technology.
In a statement released via the APUS official account, the company emphasized that the open-source initiative aims to facilitate technology exchange and collaboration, encouraging more developers to participate in the optimization and innovation of large models, collectively pushing the boundaries of artificial intelligence. The birth of this open-source large model is poised to unlock new possibilities in AI research and applications, with anticipation for the emergence of innovative applications and services based on this model in the future.
【来源】https://mp.weixin.qq.com/s/x_XSfcDBL7A4zb43vDFSMg
Views: 1