Moonshot Unveils World's Largest Open AI Model

Chinese startup Moonshot launched Kimi K3 on 17th July, a 2.8 trillion-parameter open-weight artificial intelligence model that it described as the largest of its kind and said narrows the performance gap with leading U.S. systems.

The model is designed for advanced reasoning, long-horizon coding, and knowledge work, with a context window of one million tokens. Moonshot claimed Kimi K3 outperformed Anthropic's Opus 4.8 and OpenAI's GPT-5.5 and GPT-5.6 Sol on GPU kernel optimization, a technique that improves hardware efficiency. The company also said the model performed competitively with Anthropic's frontier Fable 5 when that system's safeguards were active.

Independent benchmarks supported some of the claims. Arena.ai ranked Kimi K3 first in building web interfaces, while Vals AI placed it second overall behind Fable 5 and ahead of GPT-5.6 Sol. Artificial Analysis rated its performance comparable to GPT-5.5 and Opus 4.8 on complex multi-step tasks.

Omdia chief analyst Lian Jye Su noted that Chinese models are gaining traction because they can be deployed far more cheaply than U.S. systems, though he cautioned that scale alone does not guarantee the best performance. At 2.8 trillion parameters, the model is expensive to host locally, requiring hundreds of thousands of dollars in computing equipment.

Moonshot is backed by Alibaba and Tencent and is reportedly seeking $2 billion in new funding at a $30 billion valuation ahead of a potential Hong Kong listing. Shares of rival Chinese AI firms Zhipu and Minimax fell sharply in Hong Kong trading after the announcement.

The release follows Z.ai's recent benchmark success and underscores how China's open AI ecosystem is catching up with U.S. frontier models, challenging earlier assumptions that Chinese developers lagged by months.