Key Takeaways
- Moonshot AI unveiled Kimi K3, boasting 2.8 trillion parameters and positioning it as the largest open-weight AI system available
- The system demonstrates performance levels approaching Anthropic’s leading models while surpassing multiple American alternatives in testing
- Shortly after its debut, Kimi K3 secured the top position on UC Berkeley Arena’s coding evaluation leaderboard
- Hong Kong-listed Chinese AI companies Zhipu and Minimax saw stock prices plummet 27.7% and 16.5% following the news
- The release coincides with US regulatory obstacles preventing Anthropic and OpenAI from releasing their latest systems
On Friday, Chinese artificial intelligence company Moonshot AI introduced Kimi K3, branding it as the first open-source system approaching the 3 trillion-parameter threshold.
The system contains 2.8 trillion parameters — fundamental variables acquired through training that typically serve as a proxy for model complexity and capability.
According to Moonshot, Kimi K3 excels in sophisticated reasoning tasks, extended coding operations, and professional knowledge applications. The model incorporates a 1 million-token context window, enabling it to handle substantially more data per query compared to previous generations.
Open-weight systems enable users to access, operate, and modify the core technology independently. This distinguishes Kimi K3 from proprietary systems developed by Anthropic and OpenAI, which keep parameter specifications confidential.
Performance Comparison with American AI Systems
Moonshot reports that Kimi K3 achieved competitive results against Anthropic’s Fable 5 while exceeding the performance of Opus 4.8 and multiple GPT iterations in GPU kernel optimization evaluations.
UC Berkeley’s Arena platform ranked Kimi K3 at the top for web interface development capabilities. Vals AI positioned it in second place overall, trailing only Fable 5. Artificial Analysis drew parallels between its capabilities and OpenAI’s GPT-5.5 when handling intricate, sequential operations.
Nevertheless, Moonshot conceded that the system’s aggregate performance “still trails the most powerful proprietary models” from Anthropic and OpenAI.
Hussein Abbass, a computing professor at UNSW Canberra, noted that while Kimi K3 demonstrates strength in programming tasks, “it is still unknown how competitive it is across the whole range of tasks.”
Financial Impact and China’s Expanding AI Ambitions
The announcement triggered significant sell-offs in competing Chinese AI stocks. Zhipu and Minimax experienced share price declines of 27.7% and 16.5% respectively during Hong Kong trading sessions shortly before closing.
Several analysts compared the situation to DeepSeek’s 2025 debut, which challenged prevailing assumptions regarding American AI supremacy. Kevin Xu, a technology analyst and investor, noted he was “sensing a violent market reaction to Kimi K3, similar to a DeepSeek moment.”
Ethan Mollick, a professor at the University of Pennsylvania, characterized Kimi K3 as “closest to the frontier yet” among Chinese AI systems.
Industry experts suggest Chinese models are advancing partly due to economic advantages. Lian Jye Su from Omdia explained they “can be run at a fraction of the cost that OpenAI charges its clients.”
However, deploying a 2.8 trillion-parameter system independently would require hundreds of thousands of dollars in hardware infrastructure, according to Ryan Fedasiuk from the American Enterprise Institute.
Moonshot has received backing from Alibaba and Tencent and was reportedly pursuing $2 billion in capital at a $30 billion valuation in preparation for a possible Hong Kong stock exchange listing.
The launch arrived as American government regulations have postponed public deployments from Anthropic and OpenAI, driven by fears these systems could assist malicious actors in compromising digital security.
China’s transparent AI development ecosystem is progressively narrowing the technological divide with American competitors, as companies including Moonshot, Z.ai, and MiniMax introduce capable models with reduced expenses and accelerated development cycles.


