On November 4, emails shared by many users on social platforms showed that the international version of Trae, an AI programming tool owned by ByteDance, announced that due to service adjustments, it would no longer provide access to the Claude model. The email also pointed out that with core models such as GPT-5, Gemini-2.5-Pro, Kimi-K2-0905 and DeepSeek-V3.1-Terminus, as well as TRAE’s in-depth optimization of inference quality, code intelligence and execution process, users can still enjoy the same high level of product performance as before. "We are confident about the future."
In return, the platform will also provide Pro members with 50% additional fast request quota, which means 300 new requests per month, valid until January 31, 2026. However, on platforms such as Reddit and GitHub, many Trae users expressed dissatisfaction with the adjustment to remove the Claude model and said they would consider switching to other AI programming tools of the same type.
Image source: GitHub screenshot
Blue Whale Technology’s query found that the international version of Trae currently has built-in multiple models from OpenAI, Google, xAI and other companies, including GPT-5, GPT-4.1, Gemini-2.5-Pro, Grok-4, etc. The domestic version is connected to Byte Doubao-Seed-1.6, Alibaba Qwen-3-Coder, Kimi-K2, Zhipu GLM-4.6, DeepSeek-V3.1-Terminus and other models.
Image source: Screenshot of Byte Trae international version
In January this year, Byte took the lead in releasing the international version of Trae, and launched the domestic version of the tool in March, which supports multiple services such as code continuation, document generation, logic review and structural optimization. The domestic version of Trae has not yet launched a paid plan. The international version Pro membership costs 10 US dollars per month, with a discounted price of 3 US dollars for the first month, and can get 600 fast requests for advanced models and unlimited slow requests. According to Hong Dingkun, Vice President of Byte Technology, in June, Trae's overall monthly activity has exceeded 1 million, and more than 80% of Byte's internal engineers are using Trae to assist development.
Similar to Byte, Tencent also launched internal testing of the international version of its AI programming tool CodeBuddy IDE in July this year, and officially opened the domestic version for public testing in August. Sources close to Tencent told Blue Whale Technology that the international version of CodeBuddy had removed the Claude model before October 1, and “no significant impact has been seen so far.”
Blue Whale Technology inquired about CodeBuddy and found that the international version now has built-in OpenAI GPT-5, GPT-4o-mini, Google Gemini-2.5-Pro and other models, while the domestic version has built-in models such as Tencent Hunyuan and DeepSeek-V3. According to the aforementioned person, GLM-4.6 and Kimi-K2 are also connected internally, and the results are good. It said that Claude’s relevant adjustments will help promote the development of domestic models. In the AI programming scenario, large domestic models such as Hunyuan, DeepSeek, and Zhipu are making rapid progress and will continue to improve with technology iteration.
Image source: Screenshot of Tencent CodeBuddy International Version
AI native IDE (integrated development environment) has become a new track for major domestic Internet companies to compete. In August, Alibaba launched Qoder, an agentic programming platform for global users, saying it integrated the world's top programming models and would automatically select the optimal model. Baidu also released Wenxin Kuaicode’s independent AI native development environment tool “Comate AI IDE” in June. It does not differentiate between domestic and foreign versions and supports self-selected models from DeepSeek, Kimi, and Zhipu.
Image source: Baidu Comate AI IDE screenshot
Blue Whale Technology has learned that the adjustment of the built-in model of its AI programming tools by a major manufacturer may be related to the update of the terms of service of Anthropic, the company behind Claude. In early September, Anthropic announced that it would further strengthen regional restrictions. Any company that is directly or indirectly controlled by China and other restricted regions and holds more than 50% of the shares is prohibited from accessing Anthropic services, regardless of where the actual operations are.
After Anthropic announced the tightening of restrictions, many domestic manufacturers such as Byte Volcano Engine, SenseTime, and JD Cloud took advantage of the trend to launch migration plans. Take the "Claude API User Special Relocation Plan" provided by Zhipu as an example. Developers only need to replace the API URL to seamlessly switch from Claude to the GLM model API. Zhipu also gives away 20 million Tokens for free trial to new users, and provides developers with an exclusive monthly package for GLM-4.5 encoding, emphasizing that "the price is only 1/7 of Claude, the usage is increased by 3 times, and the speed is faster."
Anthropic is one of OpenAI’s main competitors and was founded by former OpenAI employees such as Dario Amodei. Its investors include Amazon, Google and other giants, and the company's latest valuation has reached US$183 billion. Anthropic's Claude series of models are known for their programming capabilities. In the SWE‑bench Verified evaluation set, which measures the software engineering capabilities of AI models, Claude 4.5 Sonnet topped the list with a score of 70.6%. The top ten also included models such as Alibaba Qwen3-Coder and Zhipu GLM-4.5.