Moonshot AI’s Kimi K3 Competes With OpenAI and Anthropic Models on Key Benchmarks

In Focus

• Kim K3 is the largest AI model in China with 2.8 trillion parameters
• The model performed better than GPT-5.5 and Opus 4.8 in general agents and coding
• Kimi K3 will cost more than comparable Chinese open-weight models

Moonshot AI has launched a new model which it claims performs better than OpenAI and Anthropic’s most advanced models on some benchmarks. With 2.8 trillion parameters, Moonshot’s Kimi K3 is the largest AI model in China.

How Kimi K3 Compares With OpenAI Models?

Moonshot said the new model performed better than OpenAI’s GPT-5.5 and Anthropic’s Claude Opus 4.8 in benchmarks like general agents and coding. The startup placed Kimi K3 behind the frontier models developed by the two American AI firms. The model also trails GPT-5.6 Sol and Claude Fable 5 in overall performance.

Moonshot introduced Kimi K3 a month after the U.S. government restricted Fable 5 and Claude Mythos access by foreigners due to security concerns. The launch shows how fast China’s open AI ecosystem is catching up with the most advanced U.S. systems.

“Despite persistent hardware/compute capacity constraints in China, K3 demonstrates that pre-training scaling, paired with architectural innovation, can still deliver step-change gains for flagship Chinese models,” Bank of America Analysts said in a note led by Alex Liu as per CNBC.

Kimi K3 debuted after OpenAI expanded access to GPT-5.6 Sol after the U.S. government approved its public release. Access to Anthropic’s flagship Mythos 5 model remains restricted to selected government agencies and companies.

Impact on the AI-Related Stocks

The Kimi K3’s launch affected AI-related stocks with Z.ai shares plunging by up to 30% in Hong Kong. MiniMax Group stock dropped 16% while Alibaba lost 4%. Bloomberg’s Asian semiconductor index slid more than 6%, and Nasdaq 100 futures declined 2%.

Arena blind tests ranked Kimi K3 ahead of leading U.S. models for front-end coding tasks and placed it alongside GPT-5.6 Sol in general text performance. Moonshot priced Kimi K3 at $15 per 1 million output tokens. This places the cost at about $0.94 per task, which is slightly lower than GPT-5.6 Sol’s $1.04, and half of Claude Opus 4.8’s $1.80.

However, the Kimi K3 costs more than comparable Chinese open-weight models. Moonshot AI plans to release Kimi K3’s open-weight model on July 27. This means developers cannot self-host or modify the model yet. The AI startup said the new model is available via Kimi.com, mobile apps, and its API.

What Does Kimi K3 Launch Means for the AI Race?

Kimi K3 marks a new phase in the global AI race. Its debut shows that Chinese startups can increasingly compete with leading U.S. models on performance while offering competitive pricing. The launch is expected to intensify innovation, speed up model development, and deepen the technological rivalry between the U.S. and China, even as export controls and security concerns continue to shape the industry.