Moonshot and MiniMax step up as China’s new frontier AI labs

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.

Do you have questions about the biggest topics and trends from around the world? Get the answers with SCMP Knowledge, our new platform of curated content with explainers, FAQs, analyses and infographics brought to you by our award-winning team.

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.

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.

One AI industry executive, who requested anonymity to speak freely about start-ups in the sector, said China’s deep talent pool was a key reason companies like Moonshot and MiniMax could punch above their weight, despite having far fewer high-end chips and less funding than many of their US peers.

Kimi is an advanced chatbot developed by Moonshot AI, a Chinese technology company backed by Alibaba. Photo: Shutterstock alt=Kimi is an advanced chatbot developed by Moonshot AI, a Chinese technology company backed by Alibaba. Photo: Shutterstock>

Moonshot AI, for example, continues to train models with significantly fewer high-end graphics processing units (GPUs) than its US rivals, but remains confident about its trajectory.

In a Reddit discussion in early November, an account believed to belong to Yang was asked when the company would roll out its next-generation foundational model, K3, given OpenAI’s massive data centre ambitions. Yang replied: “Before Sam’s trillion-dollar data centre is built.”

The founder of MiniMax, Yan, a former computer vision specialist at SenseTime, is betting on multimodal models that can handle text, images and video. The company is widely seen as one of the first in China to apply mixture-of-experts (MoE) architecture at scale – an approach later adopted and popularised by DeepSeek.

The latest open-source models from Moonshot AI and MiniMax have now overtaken DeepSeek’s offerings in Artificial Analysis’ rankings, drawing growing interest from investors. MiniMax, which raised US$300 million in a strategic round in July at a valuation of more than US$4 billion, has filed for a Hong Kong listing to raise up to HK$5 billion.

Moonshot AI has also secured fresh backing, raising about US$600 million in a funding round last month led by Beijing-based venture capital firm IDG Capital and Tencent Holdings, according to start-up database ITJuzi.com.

Even so, analysts warn that these start-ups will face intense competition from deep-pocketed Big Tech players. Wang Sheng, an investor at InnoAngel Fund, said it is difficult for AI start-ups to thrive when they are head-to-head with large technology companies in the same race.

“[AI start-ups] have to either jump the gun or position themselves differently,” Wang said. “Otherwise it will be difficult for them to win the same game.”

This article originally appeared in the South China Morning Post (SCMP), the most authoritative voice reporting on China and Asia for more than a century. For more SCMP stories, please explore the SCMP app or visit the SCMP’s Facebook and Twitter pages. Copyright © 2025 South China Morning Post Publishers Ltd. All rights reserved.

Copyright (c) 2025. South China Morning Post Publishers Ltd. All rights reserved.


Continue Reading