China’s Moonshot AI, backed by e‑commerce giant Alibaba Group, today unveiled its newest artificial intelligence model, Kimi K2.5, intensifying what industry observers are calling the domestic race against rival startup DeepSeek and other global AI challengers.
Moonshot announced the launch of Kimi K2.5, its most advanced model yet, introducing a native multimodal architecture that processes text, images, and video in a single system.
Meanwhile, it is worth noting that this update reflects a surging trend of omni models, led by key players in the tech industry, such as OpenAI and Google’s Alphabet Inc.
The new version of Moonshot’s Kimi is one of the several upgrades launched over the last month. With this finding, sources noted that major AI firms in China are scrambling to get ahead of DeepSeek’s impending announcement.
Regarding its upcoming announcement, sources acknowledged that DeepSeek has been hinting at a major launch lately. Moreover, the Chinese artificial intelligence company’s research lab shared key publications from prominent team members, including its CEO, Liang Wenfeng, and code on GitHub, a premier cloud-based, Microsoft-owned platform.
In the meantime, reports revealed that Moonshot secured around $500 million in December last year from its significant supporters, including Alibaba and IDG Capital, demonstrating renewed investor interest in high-growth, technology-driven ventures. Furthermore, the company reached a $4.3 billion valuation through this deal.
On the other hand, sources with knowledge of the situation noted that Moonshot planned to release an enhanced version of its primary model at a time when demand for AI is surging. Therefore, to cope with this escalating demand, the AI startup initiated new funding rounds targeting a $5 billion valuation.
This was after key Chinese AI rivals Zhipu and MiniMax Group Inc. announced in early January 2026 the successful introduction of their initial public offerings (IPOs) on the Hong Kong Stock Exchange (HKEX). Collectively, they raised more than $1 billion in the Special Administrative Regions of China.
Following their strategic approach to operations, Moonshot, Zhipu, and MiniMax Group Inc. are ranked among the top Chinese large language model developers, a competition once called the “War of One Hundred Models.”
Nonetheless, analysts alleged that many smaller firms have struggled to implement necessary technology enhancements and secure adequate funding after DeepSeek’s R1 model reached key milestones at the start of 2025.
As competition in the tech ecosystem heated up, Zhipu launched a GLM-Image, an image generation model, in January, claiming it is China’s first state-of-the-art model fully trained on local chips without relying entirely on foreign chips. Just after this launch, Alibaba announced this week the introduction of a reasoning version of its leading proprietary model, Qwen3-Max.
Concerning Alibaba’s move, Moonshot asserted that it firmly believes its K2.5 AI model will outperform its open-source rivals in several benchmark tests. Additionally, the firm noted that this improved version of its main model is narrowing the coding skills gap compared to major proprietary models.
To demonstrate its commitment to solidify its position as a leader in the industry, Moonshot made clear its intention to introduce an automated coding tool with the capabilities to compete with Claude Code, an AI-powered agentic coding tool developed by Anthropic.
Notably, Moonshot was established in March 2023 by Yang Zhilin, a former AI project lead at tech giants Meta Platforms Inc. and Google. He is also regarded as a leading expert in large language model development in China.
The artificial intelligence startup offers various subscription tiers for its chatbot and provides specialized technology solutions to enterprise clients; however, it still lags behind rivals such as Zhipu and MiniMax in commercialization.
Sharpen your strategy with mentorship + daily ideas - 30 days free access to our trading program