Kimi K3: Moonshot Unveils World’s Largest Open-Weight AI Model
Chinese AI startup Moonshot launches its 2.8 trillion-parameter Kimi K3 model, claiming frontier-level performance as competition with leading US AI companies intensifies.
Moonshot introduces Kimi K3, the world's largest open-weight artificial intelligence model.
Kimi K3 has entered the global artificial intelligence race after Chinese startup Moonshot unveiled what it describes as the world’s largest open-weight AI model, featuring 2.8 trillion parameters and performance that rivals some of the most advanced systems developed in the United States.
The launch comes just one month after Anthropic’s Fable and Mythos models were withdrawn by the US government over security concerns, highlighting the rapid pace at which Chinese AI companies are closing the technology gap with their American competitors.
Moonshot said Kimi K3 is the first open-weight AI model to approach the three-trillion-parameter milestone. The company designed the model for advanced reasoning, long-horizon coding and complex knowledge-based tasks.
The system also features a one million-token context window, allowing it to process and retain significantly more information within a single prompt than previous generations of AI models.
According to Moonshot, Kimi K3 delivered competitive performance against Anthropic’s Fable 5 while outperforming OpenAI’s Opus 4.8, GPT-5.6 Sol and GPT-5.5 in GPU kernel optimisation, a technique that improves hardware efficiency and reduces processing delays.
Independent benchmark results also strengthened the company’s claims.
Arena.ai ranked Kimi K3 first for web interface-building capabilities, while Vals AI placed it second overall behind Fable 5 and ahead of GPT-5.6 Sol. Artificial Analysis reported that the model performed at a level comparable to OpenAI’s GPT-5.5 and Anthropic’s Claude Opus 4.8, particularly on complex reasoning and multi-step problem-solving tasks.
The announcement immediately affected China’s AI market. Shares of domestic AI developers Zhipu and MiniMax fell sharply in Hong Kong trading, declining 21.9% and 13.8%, respectively, as investors reacted to Moonshot’s latest breakthrough.
SpaceX: Starship Launch Scrubbed After Engine Abort Delays 13th Test Flight
China’s AI industry has accelerated model development in recent months as competition with US technology companies intensifies.
The shift gained momentum after Z.ai introduced GLM-5.2, a model that achieved benchmark results close to leading US proprietary systems and challenged the belief that Chinese developers significantly lag behind their American counterparts.
MiniMax is also preparing to launch its own 2.7 trillion-parameter AI model during the third quarter of 2026. The company is simultaneously developing its next-generation multimodal system, H3, as competition among Chinese AI firms continues to intensify.
Before Kimi K3’s debut, DeepSeek’s V4-Pro and Meituan’s LongCat-2.0 led China’s AI sector with 1.6 trillion parameters each, while several other domestic developers had already surpassed the one-trillion-parameter threshold.
Unlike proprietary AI systems, open-weight models allow developers and organisations to download, customise and deploy the underlying model, providing greater flexibility for research and commercial applications.
Moonshot said Kimi K3 includes two major architectural improvements that increase computing efficiency and enable the model to complete complex coding tasks with minimal human supervision.
The startup has expanded rapidly with financial backing from major Chinese technology companies, including Alibaba and Tencent.
According to Bloomberg, Moonshot is seeking $2 billion in new funding at a valuation of approximately $30 billion as it prepares for a potential listing on the Hong Kong Stock Exchange.
The launch of Kimi K3 marks another significant milestone in China’s AI industry and reflects the country’s growing ambition to compete directly with the world’s leading artificial intelligence developers.
