Chinese startup Moonshot AI has unveiled a new model that it says closes the gap with leading U.S. offerings and outperforms the best-performing systems from OpenAI and Anthropic in some benchmarks.
Kimi K3 still lags behind Anthropic’s Claude Fable 5 and OpenAI’s GPT 5.6 Sol in overall performance, the company said Friday, but consistently outperformed other models tested.
According to Moonshot, the model beat Claude Opus 4.8 and GPT 5.5 – models that rank just behind the leading systems from Anthropic and OpenAI – in benchmarks including coding and general agents.
It is China’s largest AI model to date, with 2.8 trillion parameters relative to the size of its neural network.
“Despite ongoing hardware/compute capacity constraints in China, K3 shows that pre-training scaling, coupled with architectural innovation, can still make significant progress on China’s flagship models,” Bank of America analysts said in a note led by Alex Liu.
The release comes as the race for AI supremacy between the US and China heats up.
Chinese AI models are already growing in popularity among Western companies as they close the performance gap with U.S. competitors and remain cheaper to use than the most advanced offerings from American labs. U.S. lawmakers are considering how to curb the increasing adoption of Chinese AI models by domestic companies.
Table of Contents
Another DeepSeek moment?
Patrick Moorhead, CEO and principal analyst at Moor Insights and Strategy, called the market’s reaction to the new Kimi-K3 model “an overreaction that is frighteningly similar to the DeepSeek panic,” and explained in a post on
Moorhead said in the post that large language models, or LLMs, like Kimi K3 are only “accelerating the inference market and allowing it to grow faster than without it,” underscoring a broader shift in the tech sector from simply focusing on the size and assumed capabilities of a model itself to the overall application that drives the technology.
Perplexity CEO Aravind Srinivas told CNBC last week that startups and developers are more focused on finding the best ways to deploy AI models that can power their apps, rather than focusing solely on a huge, underlying system. This is one of the reasons why the freely available OpenClaw technology became so popular with developers earlier this year. This so-called harness allows programmers to more easily swap in and out different AI models that power digital assistants, allowing them to perform a range of actions without relying on a single LLM.
“The model alone is no longer the product,” Srinivas said at the time. “It’s the harness, the orchestration system, that puts the model into a very powerful harness and connects the model to a lot of tools.”
Moorhead attributes what he sees as the overreaction to Kimi K3’s release to politics, telling CNBC in an email: “There is a big debate in Washington DC about whether the US should use Chinese open source models and whether US companies should allow the Chinese to use their models.”
“The latter is ironic since the Chinese seem to be doing well with their models,” Moorhead said.
Lu Zhang, the founder and managing partner of Fusion Fund, said that despite the wide attention models like Kimi K3 can receive, most developers using the technology come “from the startup ecosystem rather than from the large corporate side.” These programmers often swap out an AI model when a more powerful version becomes available, or at least one that is cheaper and more efficient to run in their respective apps, she explained.
And while these AI models may seem extremely powerful at first glance, they are not “plug-and-play” and require a lot of technological know-how from developers to actually use their underlying capabilities, Zhang said.
Although the general discourse surrounding the open AI model space can often encompass the broader “U.S.-China competition narrative,” Zhang said there are several U.S. companies that are increasingly bringing open AI models to market, including Thinking Machines and DeepReinforce, which are backed by her venture capital fund.
It’s only a matter of time before a more advanced, open AI model captures the zeitgeist, given how quickly the entire AI field is moving, she said.
Much like the debut of DeepSeek’s R1 AI model in 2025 attracted attention because it is believed to be more cost-effective compared to proprietary technologies, the current hype surrounding Kimi K3 can be attributed to growing concerns about the overall cost and ability of AI to generate returns on investment.
Simon Koser, chief product officer at AI startup Tzafon, said Kimi K3 is rightfully impressive as it performs well in areas such as coding and can impress developers in AI labs.
“For some of these labs, cost has become a huge factor,” Koser said, emphasizing that AI leaders like Anthropic and OpenAI may be under pressure because cheaper AI models are available on the market.
Still, there are many ways to use the technology, and not every AI model excels at every task, even though initial benchmark tests may show that. Certain AI models may respond differently when they go into production than when they are tested, and there is no true all-round AI model that is superior to all others on the market.
“It’s going to look like a lot of people are changing,” Koser said. “But in practice I’m not sure the shift is that big.”
China’s AI shock
Founded in 2023 and based in Beijing, Moonshot AI is one of China’s leading model makers. Bloomberg reported that the company raised $2 billion in May at a valuation of more than $20 billion.
Backers include Chinese tech giant Alibaba, which makes the Qwen series of AI models, and Tencent.
Shares of Chinese AI competitors fell on news of the release. Z.aiwhich launched a new model amid much fanfare in June, saw its share price fall 28% on Friday. MiniMax groupanother flagship Chinese company, fell 16%.
“K3 raises the capability ceiling for Chinese AI models and shifts the burden of proof to other independent AI labs,” Liu said.
Earlier this week, Alibaba The share price rose on news that the company had entered into a partnership Apple in China. However, shares fell 4% on Friday.
“For Alibaba, while the company is benefiting from broad AI training/usage growth for its cloud service given the tight computing environment, Alibaba Qwen’s ‘open source leader’ narrative may face some challenges,” Liu said.
Choose CNBC as your preferred source on Google and never miss a moment from the most trusted name in business news.
https://www.cnbc.com/2026/07/17/moonshot-ai-kimi-k3-model-openai-anthropic-china.html
