Moonshot AI has unveiled Kimi K3, a new frontier artificial intelligence model that the Chinese startup says can compete with, and in some areas outperform, leading models from OpenAI and Anthropic.
The launch marks a significant milestone for China’s AI sector, with Kimi K3 featuring an estimated 2.8 trillion parameters, placing it among the largest and most capable AI models announced to date.
According to Moonshot AI, Kimi K3 delivers state of the art performance across coding, reasoning and autonomous agent tasks. The company highlighted new architectural innovations, including Kimi Delta Attention and Attention Residuals, which are designed to improve efficiency and model performance.
Early benchmark results suggest the model is already making an impact. Kimi K3 has claimed the top position on the Frontend Code Arena leaderboard, surpassing Anthropic’s Claude Fable 5 in coding performance. It also recorded a score of 88.3 on Terminal Bench 2.1, placing it just behind OpenAI’s GPT-5.6 Sol, which scored 88.8.
The results position Kimi K3 among the world’s strongest publicly available AI systems and underscore the rapid pace of development coming from Chinese AI companies.
Open weights planned
Moonshot AI also announced plans to release Kimi K3’s model weights by July 27, allowing developers and researchers to run, fine tune and build applications on the model locally.
If released as planned, Kimi K3 would become one of the most powerful openly available frontier models, providing an alternative to proprietary offerings from OpenAI, Anthropic and Google.
The decision could accelerate AI research and enterprise adoption while also renewing discussions around model safety, responsible deployment and international export controls for advanced AI technologies.
Industry observers say the release reflects the increasingly competitive global AI landscape, where Chinese developers are narrowing the performance gap with established Western AI laboratories.
For developers, Kimi K3’s strong coding capabilities could improve software development productivity, particularly in frontend engineering and agent based workflows. Enterprises may also benefit from another high performance foundation model that can be deployed outside commercial API ecosystems.
As investment in frontier AI continues to accelerate worldwide, Kimi K3 represents another sign that leadership in artificial intelligence is becoming increasingly global, with innovation now emerging from multiple regions rather than a small group of US companies.