World's Most Powerful Open-Source AI Model Arrives! China's Moonshot AI Unveils Kimi K3 with 2.8 Trillion Parameters

In the global race for artificial intelligence (AI) dominance, China has reached a new milestone. AI startup Moonshot AI officially launched its AI model Kimi K3 last Friday. This model is currently the world's largest open-source AI model in terms of parameter count, with performance approaching that of top international AI systems.

In the early hours of July 17, Moonshot AI officially announced the release of Kimi K3, the most powerful large model the company has ever developed. In its statement, the company said Kimi K3 features 2.8 trillion parameters and supports a context length of 1 million tokens.

Parameter count is a key metric for measuring the performance and complexity of AI models. Among global open-source models, Kimi K3 is the first to exceed 2 trillion parameters, making it the largest open-source model worldwide.

At the launch event, Moonshot AI cited test results showing that Kimi K3 outperforms models such as Anthropic's Claude Opus 4.8 and OpenAI's GPT 5.5 in programming tasks and complex applications. These test results have not yet been independently verified.

At the same time, the company acknowledged that the new system still slightly lags behind top-tier U.S. closed models—Claude Fable 5 and OpenAI's latest GPT-5.6 Sol—but the gap is rapidly closing.

This technological breakthrough could bring substantial economic returns to Moonshot AI. The company reached a valuation of $31.5 billion in its most recent funding round. Reports suggest the startup is preparing for a Hong Kong IPO to raise more capital for the competitive race.

Kimi K3 is scheduled to be freely opened to global developers by the end of July 2026. Chinese AI models are increasingly favored worldwide, partly because many are offered as free open-source software with operating costs typically lower than their Western counterparts.

Fierce Competition and Accusations of Imitation

The release of Kimi K3 comes amid heightened geopolitical tensions, reaffirming the strong momentum of Chinese AI firms striving to catch up with their U.S. peers. Observers have begun comparing this moment to the 'DeepSeek moment' of early 2025.

According to The Wall Street Journal, experts believe the U.S. leads Chinese labs by only 'six to nine months' in AI model development.

Meanwhile, serious accusations have emerged from Silicon Valley. Anthropic has accused several Chinese labs, including Moonshot AI and Zhipu, of using a technique called 'distillation' to train their own products using U.S. models.

The U.S. government views this practice as hostile and has announced it will crack down. For security reasons, the U.S. has even delayed the release of its latest AI models. The Trump administration has also attempted to slow China's development pace through strict export controls on high-performance microchips—critical components for training AI.

Xi Jinping Calls for an End to U.S. Monopoly

Beijing's leadership is leveraging technological advances to send political messages.

On the day Kimi K3 was launched, Chinese President Xi Jinping presided over the World Artificial Intelligence Conference in Shanghai and delivered his first keynote speech. Xi likened the significance of AI to the invention of the steam engine and electricity, envisioning a future where China helps the Global South build AI while leading the establishment of global AI standards.

On July 16, the World Artificial Intelligence Cooperation Organization (WAICO), initiated by China, was officially established in Shanghai. Representatives from 29 countries, including Kazakhstan, Laos, Pakistan, Russia, and Indonesia, signed the agreement to become founding members.