Chinese artificial intelligence startup, Moonshot AI, has released a new open-source model known as Kimi K2, aiming to reclaim its position in China’s increasingly competitive AI sector. According to the company, Kimi K2 boasts enhanced coding capabilities and improved general agent functions, allowing for more efficient execution of complex tasks. Moonshot claimed the model performs better than several mainstream open-source alternatives, such as DeepSeek’s V3, and rivals top U.S. models from companies like Anthropic in certain areas like programming.
The move comes amid a broader trend among Chinese tech firms to embrace open-source strategies in contrast to American giants such as OpenAI and Google, which typically keep their most powerful AI models proprietary. However, some U.S. firms, notably Meta Platforms, have also adopted open-sourcing. Industry analysts suggest this shift enables Chinese developers to demonstrate their technological edge, attract global developer communities, and resist U.S. attempts to curb Beijing’s progress in advanced technologies.
Several major Chinese companies including DeepSeek, Alibaba, Tencent, and Baidu have also released open-source AI models recently, intensifying competition in the space. Moonshot, established in 2023 by Tsinghua University graduate Yang Zhilin and backed by Alibaba, initially rose to prominence in 2024 for its long-text analysis and AI search capabilities. At its peak, its Kimi app ranked third in monthly active users by August 2024.
Despite its early success, Moonshot’s influence has waned in recent months following the emergence of low-cost models from rivals like DeepSeek, especially its disruptive R1 model released in January 2025. As a result, Moonshot’s Kimi application fell to seventh place in user rankings by June, according to data from Chinese AI tracking site aicpb.com. The Kimi K2 launch signals Moonshot’s intent to revive its competitive edge through open innovation and technical excellence.