正文：
近日，零一万物宣布其Yi API新增了Function Call功能，这一新功能为大模型与外部工具API的连接提供了更高效的手段。Yi-Large-FC模型经过特殊训练，具备良好的深度理解能力和强大的指令遵循能力，能够根据用户输入判断何时应调用函数，并使用JSON格式进行响应。该功能还完美兼容OpenAI的接口设计，能够“平替”GPT，为用户提供更贴近工具函数定义的输出。
Yi-Large-FC在权威评测榜单Berkeley Function-Calling Leaderboard中排名世界第六，优于众多海外知名模型，显示出其强大的功能调用能力。在更偏向实际调用API场景的Exec摘要得分排行中，Yi-Large-FC仅次于Claude 3.5 Sonnet，处于世界第二，显示了其在执行复杂调用任务时的精准和高效。
这一新功能的推出，对于需要实时适应快速变化世界的大模型来说至关重要。在专用任务处理、自动化工作流程以及增强模型能力等方面，Yi-Large-FC都能发挥重要作用。例如，它可以实时获取新闻、天气和股市动态，处理数据分析、图像生成、代码执行和信息检索等任务，以及自动化生成报告、处理数据和进行复杂计算等。
零一万物表示，Yi系列大模型因其出色的性能、丰富的工具链和极具竞争力的价格，已成为众多企业在中文环境下探索新业务和验证AI-first产品的理想选择。在全球范围内，Yi系列大模型已吸引了众多头部付费企业客户，涉及AI写作、AI编程、医疗、物流、消费3C、生化环材等多个领域。
知料科技创始合伙人、济南大学人工智能研究院副院长张世光教授表示，Yi系列大模型在内容生成质量、生成速度、复杂指令理解以及性价比方面都表现出色，是他们的最佳选择。此外，凭借开源多智能体框架爆火的MetaGPT也选择了接入Yi-Large模型，显示出其在模型性能和服务质量方面的竞争力。
零一万物此次推出的Yi API新功能，无疑将进一步提升大模型与工具API连接的效率，为用户提供更强大的AI服务。随着技术的不断进步，未来大模型将能够更好地服务于各个行业，推动人工智能的应用和发展。
英语如下：
News Title: “Zero One万物发布Yi API新功能：具备‘平替GPT’能力，全球排名第六”
Keywords: Yi API, Function Call, AI Agent
News Content:
Title: Zero One万物 unveils new Yi API functionality, boosting efficiency of large models and tool API connectivity
Recently, Zero One万物 announced the addition of a Function Call feature to its Yi API, offering a more efficient means for connecting large models with external tool APIs. The Yi-Large-FC model, which has been specially trained, exhibits excellent depth understanding capabilities and strong ability to follow instructions, capable of determining when to call a function based on user input and responding in JSON format. This feature also perfectly integrates with the interface design of OpenAI, capable of “平替” (substituting equally) GPT, providing users with outputs closer to the definition of tool functions.
Yi-Large-FC ranks sixth in the world on the Berkeley Function-Calling Leaderboard, outperforming numerous well-known models overseas, showcasing its powerful function-calling capabilities. In the Exec摘要得分排行, which is more inclined towards actual API calls, Yi-Large-FC ranks second only to Claude 3.5 Sonnet, demonstrating its precision and efficiency in executing complex call tasks.
The introduction of this new functionality is crucial for large models that need to adapt in real-time to the rapidly changing world. In specialized task handling, automation of work processes, and enhancing model capabilities, Yi-Large-FC plays a significant role. For instance, it can fetch real-time news, weather, and stock market updates, handle tasks such as data analysis, image generation, code execution, and information retrieval, as well as automate report generation, data processing, and complex calculations.
Zero One万物 stated that the Yi series of large models, due to their outstanding performance, rich toolchain, and competitive pricing, have become an ideal choice for many enterprises exploring new business opportunities and validating AI-first products in the Chinese environment. Globally, the Yi series of large models have attracted a number of top paying enterprise clients across various sectors such as AI writing, AI programming, healthcare, logistics, consumer electronics, biochemistry, and environmental materials.
Professor Zhang Shiguang, Founding Partner of Zhizhi Technology and Vice President of the Artificial Intelligence Research Institute of Jinan University, stated that the Yi series of large models excel in content generation quality, generation speed, complex instruction understanding, and value for money, making them their best choice. Additionally, the open-source multi-agent framework MetaGPT, which has gained popularity due to its success, has chosen to integrate with the Yi-Large model, showcasing its competitiveness in model performance and service quality.
The new Yi API functionality unveiled by Zero One万物 will undoubtedly further enhance the efficiency of connecting large models with tool APIs, providing users with more powerful AI services. As technology continues to advance, future large models will be able to serve various industries better, driving the application and development of artificial intelligence.
【来源】https://mp.weixin.qq.com/s/tMbsox0etMNKRtadoB-jLQ
Views: 2