New Chinese AI Model Baffles Competitors

The new artificial intelligence model “Kimi K3,” launched by the Chinese startup Moonshot AI the day before yesterday, has drawn widespread admiration for its capabilities, which are close to those of the most advanced US AI systems, prompting calls in the United States to act swiftly to keep pace with this development. Since the launch of DeepSeek’s “V1” tool in December 2024, Chinese AI companies have been continuously reshaping the sector’s landscape, offering open-source models that can be downloaded and modified, as well as making them available for free. These models are partially rethinking the economic foundations of AI established by major Western companies, which rely on paid, closed models. With the new “Kimi” model, a higher level has been reached, particularly in terms of scale: it was developed using 2.8 trillion parameters (adjustable standards), roughly double the size of DeepSeek V4 Pro, launched in April, which has 1.6 trillion parameters. Kimi K3 approaches advanced AI models such as Anthropic’s “V5” and OpenAI’s “GPT-5.6 Sol” in many respects. In some rankings, Moonshot AI’s tool topped the list, particularly in application and website programming, according to the benchmark rankings set by the Arena AI platform.