Trae connects to SiliconCloud to efficiently improve developer programming experience

Written by
Silas Grey
Updated on:July-12th-2025
Recommendation

ByteDance's first AI IDE Trae and SiliconCloud have joined forces to bring developers a new programming experience.

Core content:
1. Trae AI IDE core concept and full-scenario applicability
2. Operation steps for accessing the SiliconCloud platform model
3. SiliconCloud platform model advantages and cooperation prospects

Yang Fangxian
Founder of 53AI/Most Valuable Expert of Tencent Cloud (TVP)


Trae is China's first AI-native integrated development environment (AI IDE) released by ByteDance. With the core concept of "human-machine collaboration and mutual enhancement", it has very good adaptability to all scenarios of the development process, including code completion, code understanding, bug fixing, and code generation based on natural language. It is not only a development tool, but also an all-weather development "partner", bringing developers a more efficient and high-quality development experience.

Now, Trae has pre-installed the SiliconCloud platform as a model provider, which supports users to access DeepSeek-R1 & V3, QWQ-32B, Qwen2.5-Coder and other models suitable for coding in the platform through API keys to meet users' diverse model usage needs. SiliconCloud also provides multiple model APIs such as DeepSeek-R1 distilled version model (1.5B, 7B, 8B), Qwen2.5 (7B) for free, helping developers achieve "Token freedom".

This article will introduce how to access the SiliconCloud platform model in Trae.


1. Add a model


Enter  the Trae official website ( https://www.trae.com.cn/ ), download and log in, users can add models through two adding paths.

First, in the lower right corner of the input box, open the model list and click "Add Model".


Secondly, you can also go to Avatar > Settings > Trae AI > Model Management and click "Add Model".



2. Select "Silicon Flow" as the model service provider


The Add Model window is displayed on the page. Select the service provider as "Silicon Flow". You can choose DeepSeek-R1 & V3 models, or configure other models such as QWQ-32B through custom models.



3. Fill in the API key

Click the "Get API Key" button, and Trae will open the API key configuration page of the SiliconCloud platform for you. (If you have not registered/logged in to the corresponding website, you must first complete the registration/login action, https://cloud.siliconflow.cn/account/ak). After entering the page, create a new API Key and click the key to copy it.



4. Click the "Add Model" button


Trae will call the service provider's interface to detect whether the API key is valid. The possible results are as follows:

  • If the connection is successful, the custom model will be added.

  • If the connection fails, the Add Model window displays an error message and the error log returned by the service provider. You can refer to this information to troubleshoot the problem.




Conclusion


SiliconCloud is committed to providing developers with fast response, affordable prices, complete categories, stable and smooth large-model API services. Currently, the platform has dozens of cutting-edge models including the full-blooded version of DeepSeek-R1 & V3, covering categories such as language, vector, reordering, TTS, STT, image generation, and video generation, and is preset as the default model service provider by dozens of applications including ByteDance Trae. SiliconCloud will continue to actively cooperate with major AI applications to allow users to conveniently use various large models on the platform.