Chinese AI Startup Moonshot Claims Its Latest Model Rivals Top U.S. Systems

Beijing-based Moonshot AI has released a new large language model that it says narrows the gap with leading American AI systems, even outperforming offerings from OpenAI and Anthropic on certain benchmarks. The company's Kimi K3 model, which boasts 2.8 trillion parameters - making it the largest Chinese AI model to date - trails only the most advanced systems like Anthropic's Claude Fable 5 and OpenAI's GPT 5.6 Sol in overall performance, according to a Friday announcement.

On tests measuring coding and general agent capabilities, Kimi K3 beat Claude Opus 4.8 and GPT 5.5, models that sit just below the cutting edge from Anthropic and OpenAI. Bank of America analysts led by Alex Liu noted in a research report that despite hardware and compute constraints in China, the model shows that pre-training scale combined with architectural innovation can still deliver significant improvements for flagship Chinese models. The launch comes as the U.S.-China AI rivalry heats up, with Chinese models gaining traction among Western companies due to their lower cost and narrowing performance gap. U.S. lawmakers are now weighing how to curb domestic adoption of Chinese AI.

Founded in 2023, Moonshot AI has raised $2 billion at a valuation exceeding $20 billion, with backing from tech giants Alibaba and Tencent. The news sent shares of Chinese AI rivals tumbling: Z.ai dropped 28% and MiniMax Group fell 16% on Friday. Alibaba, which saw a boost earlier this week from its Apple partnership, slipped 4% as analysts suggested the Qwen series may face new competitive pressure. "K3 raises the capability ceiling for China AI models, shifting the burden of proof to other independent AI labs," Liu added.

Source: CNBC