**科大讯飞发布讯飞星火大模型V4.0 超越GPT-4 Turbo 领跑国内AI领域**
近日，一场引领人工智能风潮的新技术发布会在北京盛大召开。科大讯飞正式对外宣布推出其最新研发的讯飞星火大模型V4.0版本，标志着国内大模型能力迈入全新高度。此次发布会，不仅公布了新模型的七大核心能力全面升级，还宣布在多个领域的人工智能应用将正式落地。
相较于先前的版本，星火大模型V4.0不仅成功在文本生成、语言理解、知识问答等方面实现对GPT-4 Turbo的全面超越，更在多模态能力上取得了突破性进展，达到后者的97%水平。值得一提的是，这一版本的模型在多领域的应用性能也取得了巨大提升，例如在医疗、教育以及商业等领域，已展现出了其极高的实用价值。据科大讯飞CEO刘庆峰透露，目前星火APP下载量已达到惊人的1.31亿次，并且已经涌现出一批深受用户喜爱的应用助手。此外，随着星火大模型的加持，部分场景下的智能硬件销量也呈现爆发式增长，月均使用次数超过四千万。
值得一提的是，星火V4.0大模型的成功发布离不开全国首个国产万卡算力集群“飞星一号”的支持。这一模型是基于完全自主可控的技术研发而成，体现了我国在人工智能领域的强大实力与不断进步的自主创新能力。随着星火大模型的持续进化与应用落地，未来必将为国内乃至全球AI领域的发展注入更多活力。
英语如下：
News Title: “iFLYTEK Spark 4.0 Model: Surpassing GPT-4 Turbo and Leading the New Era of Domestic AI”
Keywords: iFLYTEK Spark 4.0 release, surpassing GPT-4 Turbo, new height of domestic big model
News Content: **iFLYTEK Unveils Spark Big Model V4.0, Surpassing GPT-4 Turbo and Leading the Domestic AI Field**
Recently, a new technology press conference that set the trend in artificial intelligence was held in Beijing. iFLYTEK officially announced the launch of its latest developed Spark Big Model V4.0, marking a new height in domestic big model capabilities. At the conference, not only was the seven-core competency upgrade of the new model unveiled, but also the official launch of AI applications in multiple fields.
Compared with previous versions, the Spark Big Model V4.0 not only successfully surpasses GPT-4 Turbo in text generation, language understanding, knowledge answering, but also achieves breakthrough progress in multi-modal capabilities, reaching 97% of the latter’s level. It is worth mentioning that the application performance of this version in multiple fields has also been greatly improved, showing high practical value in medical, educational, and commercial fields. According to iFLYTEK CEO Liu Qingfeng, the current Spark APP download has reached an astonishing 131 million times, and a number of user-favorite application assistants have emerged. In addition, with the support of the Spark big model, sales of smart devices in some scenarios have also shown explosive growth, with more than 40 million monthly uses.
It is worth mentioning that the successful release of the Spark V4.0 big model is inseparable from the support of the first domestic ten-thousand-card computing power cluster “FeiXing No.1”. This model is developed based on fully independent technology research and development, reflecting China’s strong strength in the field of artificial intelligence and continuous improvement of independent innovation capabilities. With the continuous evolution and application of the Spark big model, it will inject more vitality into the development of the domestic and even global AI field in the future.
【来源】https://www.jiqizhixin.com/articles/2024-06-28-9
Views: 3