{"id":14898,"date":"2024-07-06T08:54:03","date_gmt":"2024-07-06T00:54:03","guid":{"rendered":"https:\/\/www.1ai.net\/?p=14898"},"modified":"2024-07-06T08:54:03","modified_gmt":"2024-07-06T00:54:03","slug":"%e6%99%ba%e8%b0%b1%e5%8f%91%e5%b8%83%e5%b9%b6%e5%bc%80%e6%ba%90%e4%bb%a3%e7%a0%81%e7%94%9f%e6%88%90%e5%a4%a7%e6%a8%a1%e5%9e%8b-codegeex-%e7%ac%ac%e5%9b%9b%e4%bb%a3%ef%bc%8c%e5%8f%b7%e7%a7%b0%e7%99%be","status":"publish","type":"post","link":"https:\/\/www.1ai.net\/en\/14898.html","title":{"rendered":"Zhipu released and open-sourced the fourth generation of CodeGeeX, a large model for code generation, claiming to have the best performance for scales below 10 billion"},"content":{"rendered":"

During the ongoing 2024 World Artificial Intelligence Conference (WAIC 2024),Zhipu<\/a> AI releases andOpen Source Code<\/a>generateLarge Model<\/a> CodeGeeX<\/a> The fourth generation of\u00a0CodeGeeX4-ALL-9B<\/strong>, which integrates code completion and generation, code Q&A, code interpreter, tool call, online search, project-level code Q&A and other capabilities. It is said to be the most powerful open-source open-source open-source open-source code with parameters below 10 billion (10B).The most powerful and versatile<\/strong>A large model of the code.<\/p>\n

It is reported that this model has greatly enhanced the code generation capability based on the language capability of GLM4. CodeGeeX4-ALL-9B single model can supportCode completion and generation, code interpreter, online search, tool invocation, repository-level long code Q&A and generation<\/strong>And other functions, covering various scenarios of programming development, supporting\u00a0300+ Programming Languages<\/strong>.<\/p>\n

\"f8ceb893-7a01-4ef4-898e-56a034b57869.jpg@s_2w_820h_588\"<\/p>\n

CodeGeeX4-ALL-9B Upgrade Support\u00a0128K context<\/strong>, so that it canHandling and utilizing longer code files<\/strong>, and even information in project code, which helps the model to have a deeper understanding of complex and detailed code. Based on longer context, CodeGeeX4-ALL-9B can handle more complex project-level tasks. When the input becomes significantly longer, it can still accurately answer the content of different code files and make modifications to the code.<\/p>\n

Officially, CodeGeeX4-ALL-9B isThe "only" code model that can implement Function Call<\/strong>It has been thoroughly tested on the Berkeley Function Calling Leaderboard, including various forms of function calls, different function call scenarios, and function call executable tests.\u00a0Exec<\/span>\u00a0The call success rate in the test set exceeded 90%.<\/p>\n

Attached open source related links:<\/span><\/p>\n