Chinese Startup Moonshot’s Kimi 3 Puts Pressure on Anthropic as China–US AI Gap Narrows

Chinese AI startup Moonshot is set to release its new Kimi 3 model, which is expected to rival or exceed the performance of Anthropic's Claude Opus 4.8. The launch signals a narrowing performance gap between Chinese and U.S. frontier AI models.
Chinese Startup Moonshot’s Kimi 3 Puts Pressure on Anthropic as China–US AI Gap Narrows

Chinese Startup Moonshot’s Kimi 3 Puts Pressure on Anthropic as China–US AI Gap Narrows
Chinese AI startup Moonshot is preparing to debut its most powerful model yet, Kimi 3, intensifying competition with U.S. frontier labs and challenging assumptions that China significantly lags in top-tier AI.

Early momentum and Kimi’s rise

Moonshot first gained attention with its Kimi K2 family, which performed strongly in the open-source market and “rank[ed] high on benchmarks,” with capabilities “not too far behind the latest frontier models.” These results positioned the company as one of a small group of Chinese labs closing in on Western leaders.

The Kimi 3 reveal

In mid-July, reports emerged that the next iteration, Kimi 3 (often referred to as Kimi K3), would be released “in the coming days” as China’s largest open‑weight AI model, with an enormous parameter count of between 2 trillion and 3 trillion. The model is “expected to perform at par with or even surpass Anthropic’s Opus 4.8,” aiming to “close the gap with closed‑source models from the likes of OpenAI and Anthropic.”

On the same day, the Financial Times reported that the startup is set to launch a model “challenging Anthropic’s lead,” with Kimi K3 “expected to exceed performance of Claude Opus 4.8,” a move framed as “a sign of narrowing gap between US and China on frontier AI.”

Funding, strategy, and industry debate

Moonshot is simultaneously raising new capital at a valuation of about $31.5 billion, after securing $2 billion at a $20 billion valuation in May, underscoring investor confidence in the Kimi roadmap.

The launch lands amid a broader industry argument over the value of paying for closed, expensive models from labs such as OpenAI and Anthropic versus adopting cheaper open‑source or open‑weight systems. Some executives now “recommend companies take cheaper open source models, like those developed by DeepSeek, Z.ai, or Moonshot, and train them for their own purposes,” reflecting growing demand for transparent, customizable alternatives.


Continue reading https://foxvector.com/stories/019f6bdf-18cb-3a6a-7379-035ffd37ee45

Write a comment