【北京，4月2日】国内科技巨头APUS与战略伙伴新旦智能近日携手开创了人工智能领域的又一里程碑，成功开源了国内首个基于千亿级MoE（Mixture of Experts）架构的大模型。这一创新成果已在知名代码托管平台GitHub上发布，为全球开发者提供了全新的研究与应用资源。
APUS-xDAN大模型4.0（MoE）不仅在技术上实现了突破，更注重普惠性和实用性。该模型能够在4090等相对低端的算力芯片上流畅运行，极大地降低了大规模模型应用的硬件门槛，使得更多企业和开发者有机会利用此类先进模型，推动人工智能技术的普及。
据官方数据，APUS-xDAN大模型4.0的综合性能表现强劲，已达到业界标杆GPT-4的90%。这一成就不仅彰显了APUS与新旦智能在AI领域的研发实力，也为行业效率提升和技术创新树立了新标杆。开源这一千亿参数的MoE架构大模型，APUS和新旦智能旨在构建一个开放、协作的AI生态系统，促进全球范围内的人工智能研究与应用发展。
APUS和新旦智能的此次合作，预示着我国在人工智能基础研究和开源社区建设方面又迈出了坚实一步，有望引领全球AI技术的普惠化进程。未来，双方将继续致力于推动AI技术的创新与应用，为构建智能未来贡献力量。
英语如下：
**News Title:** “APUS and XinDan Intelligence Create a Domestic First: Opening Source Trillion-Level MoE Large Model, Empowering Low-End Chips”
**Keywords:** APUS, Trillion MoE, Open-Source Large Model
**News Content:**
**Beijing, April 2** – Domestic tech giant APUS, in collaboration with strategic partner XinDan Intelligence, has recently set another milestone in the field of artificial intelligence by successfully open-sourcing the country’s first large model based on a trillion-scale MoE (Mixture of Experts) architecture. This innovative achievement has been published on the renowned code hosting platform GitHub, providing global developers with new research and application resources.
The APUS-xDAN Large Model 4.0 (MoE) not only breaks technological barriers but also emphasizes inclusiveness and practicality. The model can run smoothly on relatively low-end computing chips like the 4090, significantly lowering the hardware threshold for the application of large-scale models. This allows more enterprises and developers the opportunity to leverage such advanced models, fostering the普及 of artificial intelligence technology.
According to official data, the APUS-xDAN Large Model 4.0 demonstrates strong comprehensive performance, achieving 90% of the industry benchmark GPT-4. This accomplishment underscores APUS and XinDan Intelligence’s R&D capabilities in AI and sets a new benchmark for industry efficiency and innovation. By open-sourcing this trillion-parameter MoE architecture, APUS and XinDan Intelligence aim to build an open and collaborative AI ecosystem, fostering global AI research and application development.
This collaboration between APUS and XinDan Intelligence signifies a robust step forward in China’s AI fundamental research and open-source community development, potentially leading the global progression towards more inclusive AI technology. In the future, the two companies will continue to dedicate themselves to driving AI innovation and application, contributing to the construction of a smarter future.
【来源】https://mp.weixin.qq.com/s/x_XSfcDBL7A4zb43vDFSMg
Views: 1