Chinese artificial intelligence start-ups Moonshot AI and MiniMax have emerged as China’s strongest contenders to rival US frontier labs in 2025 – even as DeepSeek has stolen the spotlight as the poster child for the country’s AI ambitions.
Moonshot AI, founded by 33-year-old Yang Zhilin, has sharply raised its profile in China’s AI ecosystem with the launch earlier this month of Kimi K2 Thinking, an upgraded reasoning model.
The system outperformed OpenAI’s GPT-5 and Anthropic’s Claude Sonnet 4.5 – two of the world’s most advanced closed-source AI models – on several benchmarks.
Advertisement
Deedy Das of Menlo Ventures described the release as “a turning point in AI”, noting that a Chinese open-source model had taken the top spot. Nathan Lambert of the Allen Institute for AI also praised Kimi K2 Thinking for narrowing the gap between open-source models and the world’s leading closed-source systems.
MiniMax, led by founder Yan Junjie, has also roared back onto the global AI map with the launch of its M2 model, which last month climbed to the top of a prominent leaderboard for open models.
Advertisement
MiniMax M2 achieved a record score for an open model on Artificial Analysis’s overall intelligence index, placing it ahead of Google DeepMind’s Gemini 2.5 Pro and just behind the latest US models from OpenAI and Anthropic.
The rise of these two start-ups underlines China’s potential to challenge the US in building fundamental AI models. As tech giants such as Alibaba Group Holding and ByteDance race ahead with their own large language models and AI infrastructure, a new cohort of nimble Chinese start-ups is also making steady progress to stay competitive in the global AI race. Alibaba owns the South China Morning Post.