The AI industry is witnessing a close-to-DeepSeek-2.0 moment: China has released two AI models, two major moves in the last 10 days. Moonshot AI released Kimi K3 last week. Then, Alibaba released Qwen 3.8. The best part is that both models are free, open-source, and designed to compete directly with the most advanced models from OpenAI and Anthropic. That is why these have become the talk of the town in Washington and worldwide.
Moonshot AI: Kimi K3
As shared by the company, Kimi K3 is a large language model, and it ranks above almost every US system except OpenAI’s GPT-T5.6 and Anthropic’s Claude Fable 5.
Some of its key features include that it is free and open source; hence, developers can download, modify, and build on it with no API fees. The benchmarks show it matches US models on math, coding, and complex reasoning. The AI tool is designed for handling large documents and multi-turn conversations. Moonshot AI claims it is less filtered than US counterparts, though it still carries China’s standard state-level content rules.
Alibaba: Qwen 3.8
Alibaba just launched the next-generation Qwen model. The company calls it second only to Fable 5 in capability. As already mentioned, it is free and open source, plus it holds strong performance on text, code, and vision tasks. Alibaba is pushing Qwen3.8 for business automation, e-commerce, and cloud. This AI tool is designed with less compute than comparable US models, making it cheaper to deploy.
The Quick Comparison
OpenAI and Anthropic charge per use. China’s models cost $0 to start. In terms of access, the U.S. labs keep their best models behind APIs. You can’t see or change the code. China released the full weights. You can download and make the required changes.
For those looking for the absolute best models and the OpenAI tool structure, GPT-5.6 still leads.
Why is it Important for Washington
These platforms have been launched at a sensitive time. AI growth is a significant part of the US economy. China’s free, capable models create pressure on that story. Stocks reached highs, and investors questioned.
It also split advisors around President Trump. A former AI advisor suggested that the focus of the US should be competing by being more open. He is not in favor of labs that focus on the government to eliminate their open-source competitors.
At present, the approach at the White House is different. Many officials consider frontier AI a menace and issue in national security. They are implementing a new review process to check models before releasing. Some are taking it as a licensing system.
The Bigger Question: How did China make it happen
For a long time, China has been trying to surpass the US but is limited by access to advanced chips. Those rules seem to have loosened now. US President Trump allowed Nvidia to sell more chips in China in exchange for the US government getting a new cut.
China still has very little computer power compared to the US. There is no clarity on what hardware Mooshot used for Kimi. Distillation is one of the likely factors. This is when a company trains its model through results from another model. OpenAI and Anthropic have raised their voices against it and asked the government to act. In April, the Trump administration announced steps to limit it.
Although Kimi and another AI tool are out and free. They are as capable as the Anthropic model that the US government once restricted due to national security risk.
The AI Race in the Near Future
China’s strategy seems clear: to match or exceed US performance.
For developers, it is great news.
The cost of building AI has dropped again. Now, OpenAI and Anthropic have to prove their value beyond just model quality, through tools, reliability, and trust.
For US policymakers, this is a hard choice.
One thing is clear: the assumption that spending billions on chips and data centers guarantees a win is being tested.
