【全球科技讯】近日，人工智能领域迎来了一项重大突破，总部位于旧金山的AI公司Anthropic宣布其API引入了新的提示词缓存机制，这一创新举措大幅降低了成本并缩短了延迟。根据Anthropic的介绍，该机制能够将长提示的成本降低高达90%，并将延迟降低80%。这一功能已经在Claude 3.5 Sonnet以及Claude 3 Haiku中以beta测试版的形式开放，尽管对Claude旗下最大模型Opus的支持尚未交付。
Anthropic的这一创新源自2023年的一项研究论文，该论文允许用户在会话中保留常用的上下文信息。这意味着用户可以在与模型的对话中多次引用相同的上下文信息，而不必重复承担成本。这对于需要发送大量上下文信息的用户来说尤为重要，可以显著提高对话智能体的处理效率。
Anthropic的提示词缓存功能类似于在咖啡厅中“老样子”的便捷服务，用户只需输入一次上下文信息，之后就可以快速调用。这一功能不仅适用于对话智能体，还适用于搜索工具、代码自动补全等场景，能够大幅提高工作效率。
Anthropic的提示词缓存价格也相对较低，用户只需支付初始的缓存费用，之后每次调用成本仅为正常输入价格的十分之一。这一创新不仅提升了用户体验，也使得Anthropic的API成为行业的标配，对谷歌和OpenAI等竞争对手构成了直接挑战。
随着AI技术的不断进步，Anthropic的这一创新为AI交互效率带来了重大飞跃。这一功能的推出，不仅为用户节省了成本，也提高了工作效率，有望在未来的AI应用中发挥更大的作用。
英语如下：
News Title: “Anthropic Innovates with API: Cost Reduced by 90%, Latency Cut by 80%, Becomes Industry Standard”
Keywords: Anthropic, API Innovation, Cost Efficiency
News Content:
[Global Tech Wire] A significant breakthrough has been made in the artificial intelligence sector, with the AI company Anthropic, headquartered in San Francisco, announcing the introduction of a new prompt caching mechanism to its API. This innovative approach has significantly reduced costs and shortened delays. According to Anthropic, the mechanism can slash the cost of long prompts by up to 90% and reduce latency by 80%. This feature has been made available in beta testing form for Claude 3.5 Sonnet and Claude 3 Haiku, though support for the largest model in the Claude family, Opus, is yet to be delivered.
The innovation from Anthropic stems from a research paper published in 2023, which allows users to retain common context information within conversations. This means users can reference the same context information multiple times in their interactions with the model without incurring repeated costs. This is particularly important for users who send a large volume of context information, significantly enhancing the efficiency of dialogue agents.
Anthropic’s prompt caching feature is akin to the convenience of ordering “just like last time” in a café, where users only need to input context information once and can quickly recall it later. This functionality is not only beneficial for dialogue agents but also for search tools, code auto-completion, and other scenarios, significantly boosting productivity.
The prompt caching service from Anthropic is also priced affordably, with users only paying a one-time cache fee and subsequent calls costing only one-tenth the normal input price. This innovation not only enhances the user experience but also makes Anthropic’s API the industry standard, posing a direct challenge to competitors like Google and OpenAI.
As AI technology continues to advance, Anthropic’s innovation represents a major leap forward in AI interaction efficiency. The introduction of this feature not only saves users costs but also boosts productivity, with the potential to play a larger role in future AI applications.
【来源】https://mp.weixin.qq.com/s/6Ns8F0qmsnVc536D8nunMA
Views: 2