「快意」大模型

「快意」大模型（KwaiYii）是快手AI团队精心研发的系列大规模语言模型，在人工智能领域展现出了令人瞩目的实力。其中，KwaiYii-13B作为该系列的重要组成部分，以其13B参数规模在多个权威的中/英文Benchmark上取得了杰出的成绩。KwaiYii-13B不仅拥有强大的预训练模型能力，还包含专为对话优化的模型版本，体现了快手AI团队在AI领域的研究深度和技术实力。

KwaiYii is a series of large-scale language models carefully developed by the Kwai AI team, showing impressive strength in the field of artificial intelligence. Among them, KwaiYii-13B, as an important component of this series, has achieved outstanding results in multiple authoritative Chinese/English benchmarks with its 13B parameter scale. KwaiYii-13B not only has a strong pre training model capability, but also contains a model version optimized for dialogue, which reflects the research depth and technical strength of the Kwai AI team in the AI field.

KwaiYii-13B-Base是该系列中以通用技术为基础的预训练模型，它在多个中/英文Benchmark上展示了卓越的性能，包括MMLU、CMMLU、C-Eval和HumanEval等。这些测试结果不仅验证了KwaiYii-13B-Base的学习和理解能力，也证明了它在同尺寸模型中的领先地位。KwaiYii-13B-Chat则专注于对话场景，无论是内容创作、信息咨询、数学逻辑问题解决、代码编写还是持续对话，均显示出出色的语言处理能力。与市面上的主流开源模型相比，KwaiYii-13B-Chat在多项任务中的表现接近或超过ChatGPT(3.5)的水平，展现出其在实际应用中的巨大潜力。

KwaiYii-13B Base is a pre trained model based on general techniques in this series, which has demonstrated excellent performance on multiple Chinese/English benchmarks, including MMLU, CMMLU, C-Eval, and HumanEval. These test results not only validate the learning and understanding ability of KwaiYii-13B Base, but also demonstrate its leading position in models of the same size. KwaiYii-13B Chat focuses on dialogue scenarios, demonstrating excellent language processing abilities in content creation, information consulting, mathematical and logical problem-solving, code writing, and continuous dialogue. Compared with mainstream open source models on the market, KwaiYii-13B Chat performs close to or exceeds the level of ChatGPT (3.5) in multiple tasks, demonstrating its enormous potential in practical applications.

在Benchmark评测中，KwaiYii-13B系列模型参与了行业内公认权威的Benchmark，体现了其通用性和专业性。C-Eval作为中文领域里最有影响力的综合性考试评测集之一，对模型的中文理解能力提出了严苛的挑战，而KwaiYii-13B系列模型在这样的评测中也取得了优异的成绩。如此多维度的评估显示了「快意」大模型在不同学科知识点和难度级别上的广泛适用性，为行业提供了新的标杆。

In the benchmark evaluation, the KwaiYii-13B series model participated in the industry’s recognized and authoritative benchmark, demonstrating its universality and professionalism. As one of the most influential comprehensive exam evaluation sets in the Chinese language field, C-Eval poses a rigorous challenge to the model’s Chinese understanding ability, and the KwaiYii-13B series models have also achieved excellent results in such evaluations. The multidimensional evaluation demonstrates the widespread applicability of the “Kuaiyi” model in different disciplinary knowledge points and difficulty levels, providing a new benchmark for the industry.

打不开?

• 建议用手机浏览器打开。微信/QQ可能屏蔽了该网站，首先保证网址是从浏览器/手机浏览器打开的，因为微信/QQ会屏蔽一些站。
• 建议使用不会屏蔽网址的浏览器。如果浏览器提示该网站违规，并非真的违规，而是浏览器厂商屏蔽了这个站。推荐原生态不会屏蔽网站的浏览器，苹果可以用自带的浏览器；安卓可以看以下推荐：

• 通常打不开都是因为网络问题。大型网站会针对三大运营商(电信、移动、联通)进行优化，小网站经常会遇到一些运营商问题打不开。一劳永逸的话，我们推荐使用加速器（将自己的网络切换成更稳定的运营商，比如电信）。部分境外网站需要魔法上网，比如ChatGPT和谷歌等。
• 以上三点均能解决99.99%网站打不开的问题。如有疑问，可在线留言，着急的话也可以加QQ社群，在群里@群主或管理员处理。