Chinese Startup Moonshot AI Unveils Kimi K3 Model Challenging US AI Leaders

Chinese startup Moonshot AI revealed its new AI model, Kimi K3, on Friday, claiming it narrows the performance gap with leading American AI systems and surpasses some of the most advanced models from OpenAI and Anthropic in specific benchmarks. Although Kimi K3 still trails Anthropic's Claude 5 and OpenAI's GPT 5.6 Sol in overall performance, it outperformed other models such as Claude Opus 4.8 and GPT 5.5 in programming and general agent tasks. The model is the largest Chinese AI model to date, featuring 2.8 trillion parameters, highlighting its network size.

Analysts from Bank of America, led by Alex Liu, noted that despite ongoing hardware and computational limitations in China, K3 demonstrates that scaling up pre-training combined with architectural innovation can yield significant improvements in leading AI models. This launch occurs amid an intensifying AI supremacy race between the US and China. Chinese AI models are gaining traction among Western companies due to their improving performance and lower costs. In response, US Congress members are exploring ways to limit the growing adoption of Chinese AI models by domestic firms.

Founded in 2023, Moonshot AI is among China's leading AI companies. Bloomberg reported that the startup raised $2 billion in May at a valuation exceeding $20 billion. Major Chinese tech investors include Alibaba, which develops the Qwen AI model series, and Tencent. Following Moonshot AI's announcement, shares of Chinese AI competitors fell sharply: Z.ai's stock dropped 28% after launching a new model in June, and MiniMax Group's shares declined 16%. Earlier in the week, Alibaba's stock had briefly risen due to news of a partnership with Apple in China but fell 4% on Friday.