Kimi has reportedly informed investors of corporate structure adjustments as it prepares for a Hong Kong initial public offering, with the potential to complete the listing within the next six months.
On July 17th, Moonshot AI officially announced its latest and most powerful model, Kimi K3. The model boasts 2.8 trillion total parameters, a 1 million token context window, and native multi-modal capabilities, designed specifically for long-context intelligent coding and self-evolving workflows. It is currently the world's largest open-source model by parameter count.
Code Arena, a platform created by UC Berkeley researchers, employs an anonymous double-blind evaluation system, aggregating real-world feedback from millions of global users. It is considered the gold standard for assessing large language models' coding capabilities. The Kimi K3 model achieved a top score of 1679, surpassing Claude Fable 5 and GPT-5.6 Sol. Notably, Kimi K3 ranked first in six out of seven front-end development categories within the benchmark, achieving a 76% win rate in the front-end coding arena, outperforming Fable 5's 63% and GPT-5.6 Sol's 58%.
Reports indicate the launch of Moonshot AI's new Kimi model has impacted Wall Street, contributing to significant declines in technology and semiconductor stocks at the market open.
Since its founding three years ago, Kimi has raised over 37 billion yuan in cumulative funding, making it the AI large language model startup with the most publicly disclosed financing in China. Information suggests the company has initiated a new funding round, with its pre-money valuation rising to $31.5 billion (approximately over 200 billion yuan).
Comments